Astra offers an advanced bank-to-bank transfer technology for financial institutions, fintech startups, and enterprises

Astra

astra

interactive investor has teamed up with the experts at eyeQ who use artificial intelligence and their own smart machine to analyse macro conditions and generate actionable trading signals. This time it analyses the UK drug giants.

eyeQ: GSK, Astra and the battle of big pharma

One of our goals at DataStax is to enable every developer — regardless of the language they build in — to deliver AI applications to production as fast as possible. We , our database-as-a-service built on open-source , to help developers create AI applications. recently added vector search capabilities to DataStax Astra DB Apache Cassandra Cassandra is well-known to be the most powerful, scalable, and production-ready database. With the addition of vector search, Cassandra and Astra DB have become a critical foundation for building enterprise-grade Gen AI applications. But we also need to ensure that this powerful technology is accessible and easy to use for the broadest set of developers, regardless of their preferred language or skill set. Today we take a big step in that direction by providing the huge population of JavaScript developers access to the most powerful vector database in the world, through a simple API: introducing the JSON API for Astra DB. Meet the JSON API! In the JavaScript world, document databases are prominent. No surprise there, as is a native notation in JavaScript, so an ability to store and retrieve JSON documents to and from the database tremendously accelerates the development. JSON The new JSON API is designed to provide a smooth developer experience for JavaScript developers creating new AI applications. We set out to ensure that if you’re a JavaScript developer, you can fire up an instance of Astra DB and start coding right away using paradigms and frameworks you’re familiar with. Exposing Astra DB as a document database provides multiple improvements in developer experience: You think in terms of JSON objects, which gives you natural alignment with the JavaScript ecosystem. There is no data modeling step, as it’s taken care of by the database itself. You just save and retrieve documents. You can start developing quickly, and focus on the application logic rather than on what is happening on the backend. Integration with MongooseJS We also noticed that many members of the JavaScript community work with document databases through object data modeling (ODM) libraries, specifically MongooseJS is a popular framework for object modeling on top of document databases. With and , it doesn’t require an introduction. MongooseJS. over 2 million weekly NPM downloads almost 3.5 million dependent Github repositories The new JSON API for Astra DB is fully compatible with MongooseJS. This means that it takes just a couple lines of code to point MongooseJS to an Astra DB instance: // Import MongooseJS.
const mongoose = require("mongoose");
// Import the driver for Astra DB (shipped as a part of stargate.io).
const { driver } = require("stargate-mongoose");

// Tell MongooseJS to use the Astra DB driver instead of the default one.
mongoose.setDriver(driver);

// Connect to Astra DB.
await mongoose.connect(astraDbUri, {
  isAstra: true,
}); Once connected, you can use MongooseJS APIs, and Astra DB will take care of the heavy lifting of storing your documents in an efficient way, indexing them, and scaling out when needed. Vector search support Even better, when developing with MongooseJS backed by Astra DB, you get full access to Astra DB Vector, the only database designed for simultaneous search and update on distributed data and streaming workloads with ultra-low latency, as well as highly relevant vector results that eliminate redundancies. As a result, you get the ease of use and familiarity of MongooseJS, combined with the rich vector support and scalability of Astra DB. Developing AI applications in JavaScript has never been easier! Let’s take a look at a simple example of how to use Astra DB’s vector search within a MongooseJS application. In this example, we’ll create a collection of movies with their text descriptions and some other information, like title, production year, and genre. In addition, we’ll instruct MongooseJS that we want to store for the descriptions. Here is how the model definition will look: vector embeddings const Movie = mongoose.model(
  "Movie",
  new mongoose.Schema(
    {
      title: String,
      year: Number,
      genre: String,
      description: String,
      $vector: {
        type: [Number],
        validate: (vector) => vector && vector.length === 1536,
      },
    },
    {
      collectionOptions: {
        vector: {
          size: 1536,
          function: "cosine",
        },
      },
    },
  ),
); Those familiar with MongooseJS will find this to be a typical MongooseJS model, except for two additional pieces that Astra DB’s driver allows for: The special that is used to store vector embeddings. $vector field The object that tells Astra DB how to index the vector embedding field. With the model above, you can insert documents along with the embeddings: collectionOptions.vector await Movie.insert({
  title: "In the Border States",
  year: 1910,
  genre: "Drama",
  description: "In the Border States is a 1910 American drama film...",// Generate embedding for the description,
  // for example by invoking the OpenAI API.
  $vector: embedding("In the Border States is a 1910 American drama film..."),
}); Your application can now provide functionality to enter a free-form query to search movies by their descriptions. For that, you will use the same model to generate an embedding for the user’s query, and use Astra DB’s vector search to find the most relevant entries in the database: await Movie.find({})
  .sort({ $vector: { $meta: embedding("Something funny") } })
  .limit(3); Of course, in many cases, vector search on its own isn’t enough, as you might want to combine it with filtering based on other fields in the document. For example, here is how you can find relevant movies similar to the previous example, but only looking at dramas: await Movie.find({ genre: "Drama" })
  .sort({ $vector: { $meta: embedding("Criminals and detectives") } })
  .limit(3); With MongooseJS and Astra DB, you are not limited to simple CRUD operations. You can accompany them with relevancy search using vectors, or even combine the two into powerful hybrid search queries. Getting Started with the JSON API The new JSON API is currently in public preview and is available on Astra DB to anyone who wants to try it out. Follow these three simple steps to get started: Go to and create a vector database. Astra DB Once the database is active, switch to the “Connect” tab, choose “JSON API” as your preferred method, and follow the instructions. Enjoy developing! More details on how to use the JSON API can be found in the documentation. What’s next? By introducing the JSON API, our vision is clear: we want Astra DB to be the first choice for JavaScript developers building AI applications. This is just the beginning — stay tuned for further improvements and additions. Got questions, feedback, or maybe you’re just as excited as we are? Drop us a message at . json-preview@datastax.com By Val Kulichenko, DataStax Also published here.

Leveraging Astra DB's New JSON API to Build JavaScript-Powered AI Apps

Aprovechando la nueva API JSON de Astra DB para crear aplicaciones de IA basadas en JavaScript

Tirer parti de la nouvelle API JSON d'Astra DB pour créer des applications d'IA basées sur JavaScript

जावास्क्रिप्ट-संचालित एआई ऐप्स बनाने के लिए एस्ट्रा डीबी के नए JSON एपीआई का लाभ उठाना

Astra DB の新しい JSON API を活用して JavaScript ベースの AI アプリを構築

Aproveitando a nova API JSON do Astra DB para criar aplicativos de IA baseados em JavaScript

Tận dụng API JSON mới của Astra DB để xây dựng các ứng dụng AI hỗ trợ JavaScript

利用 Astra DB 的新 JSON API 构建 JavaScript 支持的 AI 应用程序

A look at two databases that have made claims to the Kubernetes native label: TiDB and DataStax Astra DB. The cloud computing revolution has inspired and benefitted from multiple interrelated trends. The availability of self-service, public cloud infrastructure has helped to drive the adoption of microservice architectures and DevOps practices, including automation and observability. The drive toward containerization and container orchestration has led to the widespread adoption of Kubernetes as an environment for managing cloud-native applications. But one of the lagging areas in this revolution has been data and data infrastructure. For too long, data has been something that has lived outside of Kubernetes, leading to a lot of extra effort and complexity for developers in deploying cloud-native applications. One oft-repeated axiom in the early years of Kubernetes was that it was not yet ready for stateful workloads. Thankfully, a major shift has been quietly underway and has reached a point of maturity. The transformation happened slowly initially, beginning with efforts to containerize existing databases. This worked relatively well in small databases that ran on a single compute node, or databases that had been designed in a cloud-native world, like Apache Cassandra and DynamoDB, but challenges remained. Over the past two to three years, a new generation of databases has emerged. These “Kubernetes native” databases have been designed from the ground up to run on this open-source orchestration system. Here, we’ll define the qualities that make a database Kubernetes native and the benefits of adopting a Kubernetes native database. To do that, we’ll look at two databases claiming the Kubernetes native label: TiDB and DataStax Astra DB. Kubernetes Native MySQL with TiDB First, let’s examine a database with a relational emphasis: (short for Titanium Database). TiDB is an open-source system built by PingCAP that provides a and a columnar database to support hybrid transactional and analytic processing (known as HTAP, for short). TiDB MySQL-compatible database As shown in Figure 1 below, TiDB has a microservice design. The TiDB query layer, TiKV MySQL databases, TiFlash columnar databases, Spark nodes, and metadata management are each deployed as scalable microservices in their clusters. This design separates compute-intensive work from storage-intensive work, as the query and database layers are independently scalable. One critical commitment the TiDB creators made was that the database only runs on Kubernetes. Is that enough to make it Kubernetes native? Let’s dig a bit deeper. First, TiDB is deployed and managed by a Kubernetes operator using (CRDs). The TiDB CRDs include the TiDBCluster, which enables you to specify the scaling and configuration of each microservice and how the database layer components use storage through Kubernetes Persistent Volumes. Additional CRDs are used to deploy monitoring tools and manage operational tasks like backup and restore. custom resources that interfaces with the default K8s scheduler to make more application-aware scheduling decisions. This emphasis on using existing Kubernetes capabilities where available is the mark of a Kubernetes native database. TiDB also has an optional scheduler extension Kubernetes Native Cassandra with DataStax Astra DB Now, look at another Kubernetes native database and note some similarities and differences. that was one of the first to claim to be cloud native, but what does it look like to deploy Cassandra in Kubernetes? is a highly scalable NoSQL database Cassandra that has been factored into microservices, as shown in Figure 2. is a version of Cassandra DataStax Astra DB Like TiDB, the database includes microservices concerned with query processing and data storage, as well as services for identity and access control, data repair, and backup/restore. The data services are particularly interesting in their use of , with Kubernetes Persistent Volumes used only for caching and object storage used for longer-term persistence. Separating compaction into its service enables this compute-intensive processing to happen in the background without affecting the performance of data services serving read and write traffic. storage Astra DB is offered as a managed service available in multiple cloud regions. Each region contains a data plane consisting of the services mentioned above, managed by a Kubernetes operator, as well as infrastructure services, including the stack for observability and for metadata management. Kube-Promethus etcd The data planes are managed by a control plane that can run in one or more clouds to manage customer accounts and databases and provision Kubernetes clusters in new regions. in which multiple user databases can share the same microservices and supporting infrastructure, lowering unit economics for smaller-scale users. One novel aspect of Astra DB is its multi-tenant architecture As users grow their applications, they can move to dedicated resources to achieve optimal performance at scale, all on a “ ” basis. pay-as-you-go Kubernetes Native Database Principles Based on our observations of TiDB and Astra DB, we can derive some ideas of what makes a database Kubernetes native. Many of these correspond to a list of principles for cloud-native data, which I described in an : earlier article First, a database broken into constituent microservices enables each service to be scaled independently. Some types of compute-intensive processing may even be scaled to zero for a true serverless solution, especially when combined with a multitenant design. Composable microservice architecture: Microservices composed of a Kubernetes native database should make maximum usage of Kubernetes APIs for managing the fundamental resources of cloud-native applications: compute resources such as , Kubernetes ingress and services for exposing network access to data and more. This includes leveraging capabilities already present in Kubernetes, such as etcd for metadata management, instead of bringing along components with duplicative functionality. Treat compute, network, and storage as commodities: StatefulSets and deployments for managing workloads, the Persistent Volume subsystem for storage Following common patterns for Kubernetes applications will yield multiple operational benefits, for example, . By default, Kubernetes itself sets a great example that databases should follow for how to be secure: using Kubernetes Secrets to distribute security credentials, only exposing ports as needed, and so on. Leverage Kubernetes best practices: exposing liveness and readiness checks on each microservice to help availability and exposing metrics via the Prometheus PromQL API for observability A Kubernetes native database should embody the Kubernetes principles of declarative management via operators and custom resources, rather than relying on legacy database management UIs and CLIs. When necessary, . The goal is a clean separation of data plane functionality (managing data) from control plane functionality (managing the database). Declarative management via operators: Kubernetes extension points, such as scheduler extensions, can be used to add application-specific behavior Databases and other data infrastructures that faithfully adopt these principles will yield benefits, including a , lower operational complexity resulting in faster time to market, and standards-compliant solutions meeting today’s high availability and security demands. high performance for optimal cost at all scales The Future of Kubernetes Native Data Infrastructure Much progress is still to be made, and it’s not limited to databases alone. Kubernetes native principles can be applied to other types of data infrastructure, including streaming, analytics, and machine learning. Kubernetes native solutions will continue to make strides in multicluster and multi-cloud deployments to scale globally and will adopt multitenancy and serverless principles for better cost optimization. Kubernetes itself has room for improvement in adding more flexibility to StatefulSets and support for multicluster federation. The key to continued progress is open collaboration. The is a highly active group of data geeks bringing together builders of data-intensive applications and the infrastructure that supports them. Data on Kubernetes Community Join us to talk about ideas like developing reusable operators that can manage multiple databases or defining a common set of CRDs for concepts like backup/restore and data loading. Together we’ll continue to push the horizon of cloud computing for the benefit of all. Learn more about Kassandra native databases and more at the digital summit on March 14, 2023. Cassandra Forward This article is based on Chapter 7, “The Kubernetes Native Database,” from the O’Reilly book “ Managing Cloud Native Data on Kubernetes ” by Jeff Carpenter and Patrick McFadin. [ By Jeff Carpenter, DataStax Jeff Carpenter has worked as a software engineer and architect in multiple industries and as a developer advocate at DataStax, helping engineers succeed with Apache Cassandra. He's involved in multiple open source projects in the Cassandra and Kubernetes ecosystems, including Stargate and K8ssandra. He is co-author of the O’Reilly books "Cassandra: The Definitive Guide" and "Managing Cloud Native Data on Kubernetes."

Kubernetes-Native Database: TiDB Vs. DataStax Astra DB

Base de datos nativa de Kubernetes: TiDB vs. Base de datos DataStax Astra

Base de données native Kubernetes : TiDB Vs. DataStax Astra DB

कुबेरनेट्स-नेटिव डेटाबेस: TiDB बनाम। डेटास्टैक्स एस्ट्रा डीबी

Kubernetes ネイティブ データベース: TiDB Vs. DataStax アストラ DB

Banco de dados nativo do Kubernetes: TiDB vs. Data Stax Astra DB

Cơ sở dữ liệu gốc Kubernetes: TiDB Vs. DataStax Astra DB

Kubernetes 原生数据库：TiDB 对比。 DataStax Astra 数据库

HOW I BECAME A LONDON STUDENT AND WENT ASTRAY

Compliance Comes to The Metaverse with Compliance Hub Astra Protocol, a legal and compliance protocol layer that provides unparalleled security to public blockchains, announced the launch of a Compliance Hub in The Sandbox virtual environment. The new Compliance Hub provides a decentralized compliance layer to metaverse users, including Know-Your-Customer (KYC) and Anti-Money-Laundering (AML) features. The protocol uses its unique technology and the experience of reputable law firms to handle on-chain compliance challenges. Astra’s new Compliance Hub seeks to make Metaverse’s developing frontier safer for all parties. To deliver the finest KYC/AML services available, it has built a unique decentralized legal network (DLN) that includes all of the top worldwide legal and audit companies. The DLN consists of 195 nations, ensuring that rules can be satisfied practically everywhere on the planet. Users in the Metaverse may go to Compliance Hub, link to its DLN, get accredited, and acquire a non-transferable NFT that enables them privileged access to otherwise inaccessible special events. This whitelisting procedure ensures that after a user’s identification has been validated, they may participate in all of the exciting metaverse events without fear of their identity being compromised. Astra’s services help consumers overcome identification problems while safeguarding them from fraudulent accounts. Jez Ali, the founder of Astra Protocol, said: The future of metaverse development will be a long-lasting and innovative mission which has already shown incredible growth and is an exciting area of Web3. By deploying the first Compliance Hub in the metaverse, we can support this growth by adding a layer of protection for users across the world. Since it boasts some of the most desired and intriguing sites, Astra Protocol picked The Sandbox for its Compliance Hub. Most people go to find new immersive experiences and methods to communicate with one another. There is fierce competition for The Sandbox locales, with just 166,464 distinct plots of land available. The Sandbox is a virtual environment built on Ethereum that allows users to create, control, and monetize online hangout areas and gaming experiences. The Rise and Rise of Metaverse in 2022 If the previous century was marked by mass manufacturing and the advent of the Internet, the twenty-first century may be remembered for the creation and spread of the virtual world, or Metaverse, which promises to be more immersive, interactive, and collaborative than the Internet has yet to achieve. Despite the grandiose expectations and promises made by many entrepreneurs experimenting in this field, it has become abundantly evident that Blockchain technology will underlie the Metaverse and help establish a sustainable environment for all its participants. While it is not unreasonable to believe that the Metaverse will influence traditional vocations or activities that are fairly commonplace now, its impact on society and how people interact with one another will be significant. Using crypto tokens issued by companies supporting these virtual connections, consumers will be able to buy digital avatars, virtual land, and even organize a party for loved ones. Artists will perform in the Metaverse, get paid in cryptocurrency, and then trade their winnings for tangible items. The extension of the Metaverse will increase the amount of value that can be unlocked, leading to a fast expansion of the global economy. Today, a fledgling form of the Metaverse exists, with digital products like Non-Fungible Tokens (NFTs) reflecting popular art and digital relics attracting investors and crypto enthusiasts alike. With major players like Facebook, or Meta as it’s now known, entering this space and boldly signalling that the Metaverse is the new future, it’s only a matter of time before other entities follow suit, resulting in an exponential expansion of the Metaverse’s boundaries and unlocking vast amounts of value previously unknown to consumers and investors. Catch all the breaking news, and Don’t forget to like the story!

Astra Protocol Brings Decentralized Compliance to Sandbox's Metaverse by Setting up Compliance Hub

46% of Americans don't have $400 in savings. Gil Akos is helping people eliminate debt and grow savings with Astra. Davis Baer : What’s Astra and how long have you been working on it? is a mobile app that helps you manage your money and set funds aside for your savings goals — automatically. You connect all of your bank and credit card accounts, input your goals, and create the rules to fund them. Astra does the rest for you! Gil Akos : Astra The origins of our company go back . My co-founder, Sam and I, were building one-off financial models for ourselves and our respective companies. We saw the impact that predictive financial tools could have across these diverse scenarios so we decided consumers should have the same power to reach their financial goals. three years Astra for iOS What’s your background? I have a graduate and postgraduate degree in Architecture (of atoms, not bits). It’s a great field because it is both creative and technical but still building mobile consumer apps for personal finance is… a bit of a hard tack. Actually, there are a lot of similarities though: the intersection of design, experience, and code; complexities that need to come together in an effective manner; and abstract models (3D vs Financial) that we can develop and deploy to achieve our objectives. My love of technology goes way back to my NES and then the Apple IIe my family was given by a friend when I was about 8. Fast forward to Architecture School at the University of Kansas (Rock Chalk!) where I got really into 3D modeling. I just thought it was so cool that I could define geometry and then parametric systems of geometry, render them, have a look, update the system, and repeat the process. I couldn’t get enough of it, so I continued studying generative and parametric systems at Columbia. I ended up learning 5 programming languages in one semester and I’ve been programming ever since. 🤓 How does that experience get you to Astra? After Columbia, I co-founded — a consultancy that offers design technology services for other creative organizations. We were fortunate enough to work with some really cool companies like Harry’s, developing high-performance design tools and organization-wide technology platforms. It was a lot of fun! Mode Lab A data-driven functional texture for the Truman razor handle | Mode Lab We were entirely bootstrapped and ran the company on cash. As anyone offering services or working freelance will tell you, revenue is lumpy and really varies month to month. We wanted to grow our business and we couldn’t find any financial tools to help, so I started building in Excel to plan for various scenarios and pull different levers to see the effect. It worked well enough that I felt everyone should have access to powerful tools like that! financial models What motivated you to get started with your company? Sam and I both saw what financial models could do and we shared a similar experience of living in New York City as young professionals. Similar to the Bay Area, the cost of living really weighs on your checking account balance from month to month. Add in some student loans and your free cash flow is really stretched. We felt that pain directly. Even in 2015 with a ton of new fintech apps out there, we still couldn’t find anything that made managing your money easier and smarter. And when we started researching the macroeconomic trends in consumer finance, we were shocked by the stats: 46% of Americans don’t have $400 saved; those under 30 have a -2% savings rate; and the average American pays $320 per year in bank fees. At that point, we decided we just HAD to work on this problem. | Data: Non-Housing Debt Balances Center for Microeconomic Data What went into building the initial product? Initially, we were just trying to learn whether we could consume financial data and render the results in a secure and personalized format. We designed a couple of apps and gathered feedback on early prototypes and click through demos. This was great for feedback but we really felt it had to make something that worked with real data, so we built a web app and integrated the service with . That brought us a couple hundred users and the detailed feedback we needed — that seeing the info was great, but what they really wanted was for the info to something. Plaid do We started building the current product in January and ran our Invision prototypes through a series of user studies. was a crucial tool early on because we were able to get unbiased feedback from strangers and enough data points to see trends in what they were telling us. We iterated on this process and began development in parallel to building an audience with . From start to finish, it took 8 months and we officially launched on August 14th. 🚀 Erlibird Ship by Product Hunt How have you attracted users and grown your company? In our early days, we sought every means to get direct feedback from potential users. We had an online survey and we even lugged $1,000 in dollar coins to TechCrunch Disrupt. “Take a survey, get a dollar!” It’s actually funny how small that pile of coins was and how we couldn’t give them all away. But it was worth it. More recently we’ve relied heavily on Ship and the community of there to grow our audience. Last month we were admitted to which also has a great existing community of engaged, tech-savvy early adopters. And we’ve used it as an opportunity to . Ship, Makers, and Startup School have been great but it’d be misleading to say that our progress so far hasn’t included plain old email. So. Many. Emails. ✉️ Makers Y Combinator’s Startup School share our progress publicly What’s your business model? The Astra for iOS app is our first true MVP. We really wanted to make sure that we had product-market fit and our mission is to empower users to save more, which is a really big problem, so the app is free to use. We plan to offer a premium subscription, like Robinhood Gold, that unlocks additional features. But now we are focused on improving the product and growing our user base, so we can continue to learn what would make Astra something everybody loves to use. What are your key growth metrics? We measure monthly active users (who are verified and can execute transfers) as our top line growth metric. Since we launched in August, we are growing our MAUs 30% week over week. We also track registered users, bank accounts connected, transfer volume, and total amount saved in a dashboard we built. We just crossed the below thresholds and are particularly proud that our initial users are on track to save $600 per year on average: 50 MAUs 200 Registered users 250 Bank accounts connected $12,000 Transfer volume $2,500 Saved Astra’s Admin Dashboard What are your goals for the future? Near-term, we will continue to build features that our users love. We are working on more transfer “action” types, like a periodic sweep and transaction roundups. And we want these features to continue to make Astra smarter. For instance, a user might want to save $20 a week on Fridays, but our financial lives are dynamic and change from week to week. In one month, a more optimal series of transfer amounts might be $20 on week one, $18 on week two, $24, then $21. That’s net $3 for the month. Might not seem like a lot but it’s ~4% and those incremental optimizations can compound over time. Long term, we want to expand from our initial focus on savings and account-to-account transfers to the other complementary aspects of our financial lives. Paying down debt and bills are the other side of the same coin and we could use more intelligent tools to manage them effectively too. What are the biggest challenges you’ve faced and the obstacles you’ve overcome? If you had to start over, what would you do differently? Developing a consumer fintech app that is secure and minimizes the friction in the sign-up process is hard as hell and it’s resource intensive. The development cycles are longer than most and there are a ton of logistical and regulatory hurdles. Dennis Mortensen’s description of “ ” rings true for us. Anti-lean startups Our biggest singular challenge has been matching technical skills to the need to iterate quickly. For the beta web app, we built last year we relied on a lot of outside talent, thinking that we could optimize our contributors for skill. I was working on our deep learning research while other team members chipped away at things I was less familiar with like React. In retrospect and ever since I believe you should always default to optimizing for speed. That meant learning full-stack dev for mobile, so I learned React Native so we could move faster and increase our rate of learning overall. Have you found anything particularly helpful or advantageous? — They are the fastest way to dig into a skill set and get to apply what you’ve learned. Udacity’s Nanodegrees — A great external framework for pushing your growth trajectory, plus the lectures are free for anyone to watch. YC Startup School — Very powerful for responding to user questions, qualifying leads, and building a help center, and their publications are super useful. Intercom — The best tactical guidebook for startup founders that I’ve found. Best! High Growth Handbook — Because of “ ” I go back to this book every 3–6 months. The Hard Thing About Hard Things The Struggle, What’s your advice for entrepreneurs who are just starting out? The most valuable, and most infrequently glamorized, qualities for a founder are persistence and balance: “Don’t Punk Out and Don’t Quit.” — Ben Horowitz I like grit too, but in my opinion, persistence and balance are what you really need. The key part that isn’t shared frequently is that you have to set yourself up to succeed — in persisting. Long hours and are often required but you need balance too. Will working 16 hour days every day allow you to keep pushing after doing that for 6 months? It took us 3 years just to get here. doing things that don’t scale Where can we go to learn more? You can learn more about Astra at our or you can check out the Astra app on the . And you can say hi to me via or . website App Store Twitter email This interview is brought to you by , a tool to schedule and automatically repeat your posts on Facebook, Instagram, Twitter, Pinterest, LinkedIn, and Google My Business OneUp

Founder Interviews: Gil Akos of Astra

Concluding Our Characterizing Biases in Cable News Study

Fifteen years ago, crypto wasn’t on the mainstream media’s radar; but today, it frequently captures the spotlight on drive-time, lunch-time, and evening news channels, drawing both seasoned investors and curious newcomers into its orbit. While it’s exciting to witness this once-niche market gaining traction as the flood of crypto headlines inundate our newsfeeds, we have to ask: can we have too much of a good thing? Taking note of every headline as the ticker goes up; and down (sometimes multiple times a day) would be akin to suffering from investor whiplash… Sensationalism vs. substance Crypto headlines often oscillate between extremes. A prime example of this is a headline that read “Ethereum price lags”; while another read “Ethereum trader strikes gold as prices soar…” From exuberant predictions around astronomical gains to dire warnings of impending collapse - and everything in between - it can be hard to make head or tail of what’s actually going on. A broader view of crypto investment takes into account the long-term fundamentals that drive crypto adoption and value, including technological innovation, adoption by institutions and mainstream users, regulatory developments, macroeconomic trends, and societal acceptance. When investors read too much into market sentiment fueled only by headlines, they risk making short-sighted investment decisions fuelled by market manipulation and short-term noise. Understanding the broader context: the crypto market cycle “Do your own research" (DYOR) is a cardinal rule in investment circles, underlining the importance of doing your own independent analysis and due diligence before making financial decisions. In the dynamic world of crypto, this mantra holds particular significance where relying solely on hearsay (including mainstream media narratives) or social media hype can lead investors astray, exposing them to unnecessary risks and potential losses. To navigate the crypto landscape effectively, astute investors are those who have taken the time to understand the nuances of Bitcoin’s market cycle (and its effect on altcoins such as Ethereum). This four-year cycle is often referred to as the ‘halving cycle’ and has been historically closely linked to price, with the cycle progressing through ‘Bull’ (rising prices); ‘Bear’ (falling prices); ‘Accumulation’ (a leveling out of prices) and ‘Expansion’ (steady growth). Importantly, within each of these phases, there might be rallies, reactions, and a lot of volatility on (hopefully) the way up. Disclaimer: Crypto is volatile, carries risk and the value can go up and down. Past performance is not an indicator of future returns. Please DO YOUR OWN RESEARCH! Back to basics: time-tested investment strategies Amidst a frenzy of post-halving headlines, investor disappointment around the absence of immediate spikes in price is real, but now is not the time to abandon proven investment principles. Instead, investors should anchor their strategies in time-tested approaches designed to weather market volatility: Dollar-cost averaging (DCA): Embrace the discipline of regularly investing fixed amounts, irrespective of market conditions. DCA mitigates the impact of short-term price fluctuations by averaging out costs over time and promoting a steady accumulation of assets.


The Hold-On-for-Dear-Life (HODL) strategy: In a market rife with volatility, the HODL mantra focuses on the importance of long-term conviction over short-term speculation. By focusing on the intrinsic value and fundamentals of assets, rather than succumbing to market hysteria, investors can cultivate their resilience.


Diversification: Diversification remains a cornerstone of sound investment practice, and sees investors spread investments across different assets (such as types of coins) to minimize risk from volatile market swings. A diversified portfolio fosters resilience and safeguards against unforeseen market turbulence. Don’t let hype eclipse substance! While an influx of crypto headlines signifies exciting maturation of the digital asset ecosystem, excessive fixation on sensational news can prove highly detrimental to investors' long-term success. By exercising discernment amidst the noise, understanding the broader market context, and adhering to time-tested investment strategies, investors can navigate the crypto landscape with confidence. Fifteen years ago, crypto wasn’t on the mainstream media’s radar; but today, it frequently captures the spotlight on drive-time, lunch-time, and evening news channels, drawing both seasoned investors and curious newcomers into its orbit. crypto While it’s exciting to witness this once-niche market gaining traction as the flood of crypto headlines inundate our newsfeeds, we have to ask: can we have too much of a good thing? Taking note of every headline as the ticker goes up; and down (sometimes multiple times a day) would be akin to suffering from investor whiplash… Sensationalism vs. substance Sensationalism vs. substance Crypto headlines often oscillate between extremes. A prime example of this is a headline that read “Ethereum price lags”; while another read “Ethereum trader strikes gold as prices soar…” From exuberant predictions around astronomical gains to dire warnings of impending collapse - and everything in between - it can be hard to make head or tail of what’s actually going on. actually A broader view of crypto investment takes into account the long-term fundamentals that drive crypto adoption and value, including technological innovation, adoption by institutions and mainstream users, regulatory developments, macroeconomic trends, and societal acceptance. When investors read too much into market sentiment fueled only by headlines, they risk making short-sighted investment decisions fuelled by market manipulation and short-term noise. Understanding the broader context: the crypto market cycle Understanding the broader context: the crypto market cycle “Do your own research" (DYOR) is a cardinal rule in investment circles, underlining the importance of doing your own independent analysis and due diligence before making financial decisions. In the dynamic world of crypto, this mantra holds particular significance where relying solely on hearsay (including mainstream media narratives) or social media hype can lead investors astray, exposing them to unnecessary risks and potential losses. To navigate the crypto landscape effectively, astute investors are those who have taken the time to understand the nuances of Bitcoin’s market cycle (and its effect on altcoins such as Ethereum). This four-year cycle is often referred to as the ‘halving cycle’ and has been historically closely linked to price, with the cycle progressing through ‘Bull’ (rising prices); ‘Bear’ (falling prices); ‘Accumulation’ (a leveling out of prices) and ‘Expansion’ (steady growth). Importantly, within each of these phases, there might be rallies, reactions, and a lot of volatility on (hopefully) the way up. Disclaimer: Crypto is volatile, carries risk and the value can go up and down. Past performance is not an indicator of future returns. Please DO YOUR OWN RESEARCH! Disclaimer: Crypto is volatile, carries risk and the value can go up and down. Past performance is not an indicator of future returns. Please DO YOUR OWN RESEARCH! Back to basics: time-tested investment strategies Back to basics: time-tested investment strategies Amidst a frenzy of post-halving headlines, investor disappointment around the absence of immediate spikes in price is real, but now is not the time to abandon proven investment principles. Instead, investors should anchor their strategies in time-tested approaches designed to weather market volatility: absence of immediate spikes in price Dollar-cost averaging (DCA): Embrace the discipline of regularly investing fixed amounts, irrespective of market conditions. DCA mitigates the impact of short-term price fluctuations by averaging out costs over time and promoting a steady accumulation of assets. The Hold-On-for-Dear-Life (HODL) strategy: In a market rife with volatility, the HODL mantra focuses on the importance of long-term conviction over short-term speculation. By focusing on the intrinsic value and fundamentals of assets, rather than succumbing to market hysteria, investors can cultivate their resilience. Diversification: Diversification remains a cornerstone of sound investment practice, and sees investors spread investments across different assets (such as types of coins) to minimize risk from volatile market swings. A diversified portfolio fosters resilience and safeguards against unforeseen market turbulence. Dollar-cost averaging (DCA): Embrace the discipline of regularly investing fixed amounts, irrespective of market conditions. DCA mitigates the impact of short-term price fluctuations by averaging out costs over time and promoting a steady accumulation of assets. Dollar-cost averaging (DCA): Embrace the discipline of regularly investing fixed amounts, irrespective of market conditions. DCA mitigates the impact of short-term price fluctuations by averaging out costs over time and promoting a steady accumulation of assets. Dollar-cost averaging (DCA): The Hold-On-for-Dear-Life (HODL) strategy: In a market rife with volatility, the HODL mantra focuses on the importance of long-term conviction over short-term speculation. By focusing on the intrinsic value and fundamentals of assets, rather than succumbing to market hysteria, investors can cultivate their resilience. The Hold-On-for-Dear-Life (HODL) strategy: In a market rife with volatility, the HODL mantra focuses on the importance of long-term conviction over short-term speculation. By focusing on the intrinsic value and fundamentals of assets, rather than succumbing to market hysteria, investors can cultivate their resilience. The Hold-On-for-Dear-Life (HODL) strategy: Diversification: Diversification remains a cornerstone of sound investment practice, and sees investors spread investments across different assets (such as types of coins) to minimize risk from volatile market swings. A diversified portfolio fosters resilience and safeguards against unforeseen market turbulence. Diversification: Diversification remains a cornerstone of sound investment practice, and sees investors spread investments across different assets (such as types of coins) to minimize risk from volatile market swings. A diversified portfolio fosters resilience and safeguards against unforeseen market turbulence. Diversification: Don’t let hype eclipse substance! While an influx of crypto headlines signifies exciting maturation of the digital asset ecosystem, excessive fixation on sensational news can prove highly detrimental to investors' long-term success. By exercising discernment amidst the noise, understanding the broader market context, and adhering to time-tested investment strategies, investors can navigate the crypto landscape with confidence.

Why Obsessing Over Crypto Headlines Is a Dangerous Investment Habit

The rise of electronic health records and the growing use of data analytics in healthcare have made data governance a key focus area for improving healthcare services to be safer, more efficient, and better tailored to individual patient needs. The stakes are high, as a single breach in healthcare data can jeopardize patient confidentiality and weaken trust in the entire healthcare system. Astra Security reports that 51% of healthcare organizations have seen an uptick in data breaches since 2019, with hospitals facing increasing amounts of breaches over the last 12 years. Data governance is important not just for security but for ensuring data quality, integrity, and accessibility – all critical for informed decision-making in healthcare. The HIPAA (Health Insurance Portability and Accountability Act) Journal revealed that a whopping 133 million records were exposed or mishandled in 2023 alone, underscoring the need for proper data governance. Healthcare data leaders like Nithin Narayan Koranchirath are rising to the challenge by championing proper data governance strategies in the healthcare industry. Nithin’s belief is that more awareness of the essential role data governance plays in healthcare will lead to increased adoption by the industry. Managing Data in Healthcare Data governance is not a box to tick and then be forgotten about; it is a strategic necessity in the healthcare industry. With the increasing amount of Electronic Health Records (EHRs) and medical research data, healthcare institutions must guarantee that data remains precise, accessible, and secure. With inconsistencies in data across different EHR systems, there are more administrative tasks and potential risks to patient well-being than ever before. The challenges for healthcare providers in handling data are maintaining privacy and security standards, complying with regulations, managing data sources effectively, and ensuring seamless communication between different systems. Establishing a data governance program can enhance the delivery of healthcare services, foster innovation and research, improve decision-making in healthcare, and ensure compliance with laws like HIPAA and GDPR (General Data Protection Regulation). An Authority on Healthcare Data Management Nithin Narayan Koranchirath, a professional in healthcare data management, stresses the significance of data governance. He has years of expertise overseeing various healthcare data systems, and emphasizes that "efficient data management goes beyond compliance, it aims at enhancing outcomes and streamlining healthcare services." Data governance might appear as a subject typically reserved exclusively for IT experts, but its impact extends to everyone involved in the healthcare network. At its core, data governance entails managing data from the collection phase to its end use in patient care or research. Nithin’s pivotal moment came when leading the Data and Analytics platform at his current organization, one of the United States’s largest healthcare providers, where he witnessed the transformative impact of data-driven insights. "This role allowed me to witness firsthand the power of data-driven insights in revolutionizing data analytics delivery and administration," Nithin reflects, solidifying his commitment to analytics and data-driven decision-making. Grasping proper data governance practices empowers healthcare professionals with the knowledge to make informed decisions. This fundamental understanding reduces the chances of mistakes that could jeopardize safety. For patients, it offers reassurance that their personal data is handled with care and confidentiality. Additionally, implementing data governance measures such as utilizing Enterprise Master Patient Indexes (EMPI) can significantly enhance healthcare quality by providing an overview of patient records. Implementing a Data Governance Framework Tailored for Healthcare Healthcare professionals frequently encounter obstacles with data, such as difficulty accessing and safeguarding data, disorganized and isolated data sets, limited resource availability, and insufficient staff collaboration and training. Overcoming these challenges involves defining roles for adopting security measures, selecting a data management solution, allocating resources for data experts, and educating staff on effective data management practices. When assessing options for data governance solutions, healthcare institutions should take into account several factors, including integration with existing systems, compliance with requirements, scalability, and ease of use. Platforms like Big Data, AWS, and Snowflake offer data governance capabilities for overseeing data governance processes while effectively addressing security concerns and regulatory demands. Nithin’s work in implementing a comprehensive data governance framework at his organization via Big Data, AWS, and Snowflake underscores the practical benefits of such knowledge. "The implementation of robust data governance practices enabled us to achieve greater compliance with regulatory requirements and mitigate risks associated with data management," he explains. For healthcare professionals, this translates to better decision-making capabilities, and for patients, it means the assurance of data privacy and security. AWS and Snowflake are cloud-based data management solutions that align seamlessly with the needs of the healthcare industry. It offers functionalities like data discovery, traceability, and search capabilities to help organizations streamline their approach to data governance while simultaneously enhancing security and compliance measures efficiently. At the end of the day, leaders such as C-suite executives and IT managers ultimately bear responsibility for ensuring data governance practices are followed in an organization. However, every individual working within the healthcare sector plays a role in its success. Effective leadership should ensure that all staff members know the importance of data governance, while data experts should have a grasp on how data impacts staff operations. Future Outlook for Data Governance in Healthcare In an era defined by numerous technological advancements in healthcare, the importance of data governance is only set to grow. As we look towards the future, data governance in healthcare is poised to become a cornerstone of the United States healthcare system, playing a pivotal role in shaping economic outcomes. The U.S. healthcare sector, one of the largest contributors to the national economy, faces rising costs, inefficiencies, and demands for improved patient outcomes and privacy. Nithin’s hard work in solving these issues will have lasting positive effects not only on the healthcare industry but on the tech sector and the U.S. economy at large. Additionally, with the standardization of data practices, interoperability across different healthcare systems can be improved. This improvement facilitates better coordination of care, especially for chronic disease management, which is a significant expenditure for the U.S. economy. Data governance addresses these challenges by enhancing the quality, accessibility, and security of healthcare data. The path toward establishing effective data governance within healthcare is challenging but critical. As Nithin Narayan Koranchirath says himself, "Data governance is the backbone of a trustworthy healthcare system." By ensuring the integrity, accessibility, and security of healthcare data, the U.S. can improve its healthcare outcomes, reduce costs, and foster innovation, thereby enhancing its economic stability and growth. This story was distributed by Jon Stojan Media under HackerNoon’s Brand As An Author Program. Learn more about the program here: https://business.hackernoon.com/brand-as-author The rise of electronic health records and the growing use of data analytics in healthcare have made data governance a key focus area for improving healthcare services to be safer, more efficient, and better tailored to individual patient needs. The stakes are high, as a single breach in healthcare data can jeopardize patient confidentiality and weaken trust in the entire healthcare system. Astra Security reports that 51% of healthcare organizations have seen an uptick in data breaches since 2019, with hospitals facing increasing amounts of breaches over the last 12 years. Data governance is important not just for security but for ensuring data quality, integrity, and accessibility – all critical for informed decision-making in healthcare. The HIPAA (Health Insurance Portability and Accountability Act) Journal revealed that a whopping 133 million records were exposed or mishandled in 2023 alone, underscoring the need for proper data governance. 51% of healthcare organizations 51% of healthcare organizations 133 million records 133 million records Healthcare data leaders like Nithin Narayan Koranchirath are rising to the challenge by championing proper data governance strategies in the healthcare industry. Nithin’s belief is that more awareness of the essential role data governance plays in healthcare will lead to increased adoption by the industry. Managing Data in Healthcare Managing Data in Healthcare Data governance is not a box to tick and then be forgotten about; it is a strategic necessity in the healthcare industry. With the increasing amount of Electronic Health Records (EHRs) and medical research data, healthcare institutions must guarantee that data remains precise, accessible, and secure. With inconsistencies in data across different EHR systems, there are more administrative tasks and potential risks to patient well-being than ever before. The challenges for healthcare providers in handling data are maintaining privacy and security standards, complying with regulations, managing data sources effectively, and ensuring seamless communication between different systems. Establishing a data governance program can enhance the delivery of healthcare services, foster innovation and research, improve decision-making in healthcare, and ensure compliance with laws like HIPAA and GDPR (General Data Protection Regulation). An Authority on Healthcare Data Management An Authority on Healthcare Data Management Nithin Narayan Koranchirath, a professional in healthcare data management, stresses the significance of data governance. He has years of expertise overseeing various healthcare data systems, and emphasizes that "efficient data management goes beyond compliance, it aims at enhancing outcomes and streamlining healthcare services." Data governance might appear as a subject typically reserved exclusively for IT experts, but its impact extends to everyone involved in the healthcare network. At its core, data governance entails managing data from the collection phase to its end use in patient care or research. Nithin’s pivotal moment came when leading the Data and Analytics platform at his current organization, one of the United States’s largest healthcare providers, where he witnessed the transformative impact of data-driven insights. "This role allowed me to witness firsthand the power of data-driven insights in revolutionizing data analytics delivery and administration," Nithin reflects, solidifying his commitment to analytics and data-driven decision-making. Grasping proper data governance practices empowers healthcare professionals with the knowledge to make informed decisions. This fundamental understanding reduces the chances of mistakes that could jeopardize safety. For patients, it offers reassurance that their personal data is handled with care and confidentiality. Additionally, implementing data governance measures such as utilizing Enterprise Master Patient Indexes (EMPI) can significantly enhance healthcare quality by providing an overview of patient records. Implementing a Data Governance Framework Tailored for Healthcare Implementing a Data Governance Framework Tailored for Healthcare Healthcare professionals frequently encounter obstacles with data, such as difficulty accessing and safeguarding data, disorganized and isolated data sets, limited resource availability, and insufficient staff collaboration and training. Overcoming these challenges involves defining roles for adopting security measures, selecting a data management solution, allocating resources for data experts, and educating staff on effective data management practices. When assessing options for data governance solutions, healthcare institutions should take into account several factors, including integration with existing systems, compliance with requirements, scalability, and ease of use. Platforms like Big Data, AWS, and Snowflake offer data governance capabilities for overseeing data governance processes while effectively addressing security concerns and regulatory demands. offer data governance capabilities offer data governance capabilities Nithin’s work in implementing a comprehensive data governance framework at his organization via Big Data, AWS, and Snowflake underscores the practical benefits of such knowledge. "The implementation of robust data governance practices enabled us to achieve greater compliance with regulatory requirements and mitigate risks associated with data management," he explains. For healthcare professionals, this translates to better decision-making capabilities, and for patients, it means the assurance of data privacy and security. AWS and Snowflake are cloud-based data management solutions that align seamlessly with the needs of the healthcare industry. It offers functionalities like data discovery, traceability, and search capabilities to help organizations streamline their approach to data governance while simultaneously enhancing security and compliance measures efficiently. At the end of the day, leaders such as C-suite executives and IT managers ultimately bear responsibility for ensuring data governance practices are followed in an organization. However, every individual working within the healthcare sector plays a role in its success. Effective leadership should ensure that all staff members know the importance of data governance, while data experts should have a grasp on how data impacts staff operations. Future Outlook for Data Governance in Healthcare Future Outlook for Data Governance in Healthcare In an era defined by numerous technological advancements in healthcare, the importance of data governance is only set to grow. As we look towards the future, data governance in healthcare is poised to become a cornerstone of the United States healthcare system, playing a pivotal role in shaping economic outcomes. The U.S. healthcare sector, one of the largest contributors to the national economy, faces rising costs, inefficiencies, and demands for improved patient outcomes and privacy. Nithin’s hard work in solving these issues will have lasting positive effects not only on the healthcare industry but on the tech sector and the U.S. economy at large. Additionally, with the standardization of data practices, interoperability across different healthcare systems can be improved. This improvement facilitates better coordination of care, especially for chronic disease management, which is a significant expenditure for the U.S. economy. Data governance addresses these challenges by enhancing the quality, accessibility, and security of healthcare data. The path toward establishing effective data governance within healthcare is challenging but critical. As Nithin Narayan Koranchirath says himself, "Data governance is the backbone of a trustworthy healthcare system." By ensuring the integrity, accessibility, and security of healthcare data, the U.S. can improve its healthcare outcomes, reduce costs, and foster innovation, thereby enhancing its economic stability and growth. This story was distributed by Jon Stojan Media under HackerNoon’s Brand As An Author Program. Learn more about the program here: https://business.hackernoon.com/brand-as-author This story was distributed by Jon Stojan Media under HackerNoon’s Brand As An Author Program. Learn more about the program here: https://business.hackernoon.com/brand-as-author This story was distributed by Jon Stojan Media under HackerNoon’s Brand As An Author Program. Learn more about the program here: https://business.hackernoon.com/brand-as-author https://business.hackernoon.com/brand-as-author

Why Data Governance in Healthcare Matters in 2024 With Nithin Narayan Koranchirath

কেন নিথিন নারায়ণ কোরানচিরথের সাথে 2024 সালে স্বাস্থ্যসেবা বিষয়ে ডেটা গভর্নেন্স

Warum Data Governance im Gesundheitswesen im Jahr 2024 wichtig ist – mit Nithin Narayan Koranchirath

Por qué es importante la gobernanza de datos en la atención sanitaria en 2024 con Nithin Narayan Koranchirath

Pourquoi la gouvernance des données dans le secteur de la santé est importante en 2024 avec Nithin Narayan Koranchirath

2024 में स्वास्थ्य सेवा में डेटा गवर्नेंस क्यों मायने रखता है, नितिन नारायण कोरानचिराथ के साथ

2024 年に医療におけるデータ ガバナンスが重要な理由 (ニティン ナラヤン コランチラート氏)

Nithin Narayan Koranchirath와 함께 2024년에 의료 분야의 데이터 거버넌스가 중요한 이유

Por que a governança de dados na área da saúde é importante em 2024 com Nithin Narayan Koranchirath

Почему управление данными в здравоохранении имеет значение в 2024 году с Нитин Нараян Коранчират

Nithin Narayan Koranchirath ile 2024'te Sağlık Hizmetlerinde Veri Yönetişimi Neden Önemli?

Tại sao quản trị dữ liệu trong lĩnh vực chăm sóc sức khỏe lại quan trọng vào năm 2024 với Nithin Narayan Koranchirath

Nithin Narayan Koranchirath 谈 2024 年医疗保健数据治理为何如此重要

This article is a collection of tips and experiences that might help you, or to which you might relate as a fellow generalist. Quick introduction — skippable I’ve been struggling with being a generalist most of my adult life. During my deepest identity and career crisis, I decided to talk to a psychologist. During our 2nd session he says, “Adam, you are most likely what we call a ‘multipotentiality’.” I go, “What? Does that mean I am like… smart or something?” “Not necessarily,” he replies, “it’s more about curiosity. People like you are more common nowadays. You should embrace it." It was the first time I became conscious of this fact. Before it was like “There is too many things. I am lost. I don’t know what to pursue. I want to pursue all at the same time.”, and it suddenly flipped to “Oh, this is who I am. This is how it’s supposed to be. I hate labeling things but,.. this label helps.” How to tell if you are a generalist? You focus on the bigger picture
You like learning
You are drawn to things you don’t know
As soon as you learn something, you consider it obvious, as if not even worth mentioning
You experience burnout quite often
You struggle with defining yourself If you managed to identify yourself, congratulations! I guess.. :/ It’s not something you can opt out of. Your head is just wired weird, unfortunately. Here are a few things you will probably need to live with: Prepare to sacrifice time Imagine mining gold. Imagine you knew exactly where the gold is. Now despite the fact that you know this, since you are a generalist, you will start digging in 27 other directions. Why? Well, exploration! Right? Something might be there, we never know. Going straight for the gold sounds too easy and boring. After excavating 27 tunnels that lead absolutely nowhere, you, naturally, start digging the 28th tunnel. You eventually end up burnout and decide to pick up the gold. That’s what I mean when I say prepare for a time sacrifice. As a generalist you need to get ready for this. It might take you longer to reach your goals, because your natural curiosity will often lead you astray. Maybe the 29th tunnel would hide something extraordinary? The urge to creativity As a generalist, you might often feel the urge to do things differently. Doing things the same as before feels mundane. But reinventing the wheel comes with a cost. Wheel is too powerful. Tested by centuries of time. Even vehicles that fly need wheels. Yet you might feel like trying and reimagining it. Here is my advice —just know when. Just know when you can afford to be innovative, and when it’s a death sentence (wheel). Anytime you feel the urge, just ask yourself “Is this a good time, or should I just force myself to do it as everyone expects me to?”. Trust me, you will get better at answering this question over time. Potential employers don’t trust you After I finished studying design, I was looking for a job either as a UX/UI designer or a frontend developer. I ended up working as a B2B marketing expert. Yeah.. Even though I can code — just launched an app — and have been practicing UX/UI design for more than 8 years, employers generally don’t trust me when they see all these skills in my CV. This is something us generalists simply have to reckon with. I recommend making multiple CVs — one for each of your skills. Also, we will have a much harder time building a strong portfolio in multiple areas, compared to a specialist, who is laser focused on a single skill. Again, it will just take more time for us.. Thrive on chaos Trying to get your life perfectly sorted out might be great for most people. Not for most generalists. But don’t worry about it. Or try not to. Continue exploring. I know it’s hard with all the looks and pressure from your family and peers who try hard to follow linear trajectories. Startups. Your perfect environment? Startups are companies that are just starting out. While their resources might not be scarce, they need to spend them wisely. If they can hire a lawyer and a data scientist in one person (I met this guy!), they will gladly do it, while a serious law firm might be like “What?!”. Also, there are not as many rules at a startup. The company is new, responsibilities are not strictly divided, rigorous processes are not yet in place. As a generalist, you probably enjoy wearing multiple hats, and startups reward you for wearing them! Don’t be afraid As a generalist, fear is your greatest enemy. People fear the unknown, but the unknown and the generalist are the very best friends. Don’t be afraid to accept tasks or jump into things you have zero experience with. This is where your adaptability shines. Jump in, adapt, figure out. You will likely have a much easier time doing this than other people would. Being broke If you are broke, the natural tendency is to find a job. With too many skills, this might prove surprisingly difficult. If you struggle, use your creativity. Try to come up with a business idea. Chances are your ideas will be above average. Use your wide range of skills to offer a unique service. I know this advice is not much. You will probably have to find the job eventually. Just know that thinking outside the box is okay even when it comes to money. Get ready to burn out I don’t know. I still struggle with this one. I guess just get ready for it. You juggle 10 balls cause you have to. Juggling 3 doesn’t quite stimulate you, you will end up bored and you will procrastinate. Too much = burnout, too little = boredom and procrastination. If anyone has tips on how to find a balance here, please let me know. To sum it up: Your strengths: Extremely adaptable


Quickly acquire new skills


Eat challenges Your weaknesses: Huge time sacrifice


Identity struggles


Burnout Key takeaways: Don’t resist, embrace
Prepare for the long run
Know when to be innovative
Don’t fear the challenge
Try balance burnout and boredom This article is a collection of tips and experiences that might help you, or to which you might relate as a fellow generalist. Quick introduction — skippable I’ve been struggling with being a generalist most of my adult life. During my deepest identity and career crisis, I decided to talk to a psychologist. During our 2nd session he says, “Adam, you are most likely what we call a ‘multipotentiality’.” I go, “What? Does that mean I am like… smart or something?” “Not necessarily,” he replies, “it’s more about curiosity. People like you are more common nowadays. You should embrace it." It was the first time I became conscious of this fact. Before it was like “There is too many things. I am lost. I don’t know what to pursue. I want to pursue all at the same time.”, and it suddenly flipped to “Oh, this is who I am. This is how it’s supposed to be. I hate labeling things but,.. this label helps.” How to tell if you are a generalist? How to tell if you are a generalist? You focus on the bigger picture You like learning You are drawn to things you don’t know As soon as you learn something, you consider it obvious, as if not even worth mentioning You experience burnout quite often You struggle with defining yourself You focus on the bigger picture You like learning You are drawn to things you don’t know As soon as you learn something, you consider it obvious, as if not even worth mentioning You experience burnout quite often You struggle with defining yourself If you managed to identify yourself, congratulations! I guess.. :/ It’s not something you can opt out of. Your head is just wired weird, unfortunately. Here are a few things you will probably need to live with: Prepare to sacrifice time Imagine mining gold. Imagine you knew exactly where the gold is. Now despite the fact that you know this, since you are a generalist, you will start digging in 27 other directions. Why? Well, exploration! Right? Something might be there, we never know. Going straight for the gold sounds too easy and boring. After excavating 27 tunnels that lead absolutely nowhere, you, naturally, start digging the 28th tunnel. You eventually end up burnout and decide to pick up the gold. That’s what I mean when I say prepare for a time sacrifice. As a generalist you need to get ready for this. It might take you longer to reach your goals, because your natural curiosity will often lead you astray. Maybe the 29th tunnel would hide something extraordinary? The urge to creativity As a generalist, you might often feel the urge to do things differently. Doing things the same as before feels mundane. But reinventing the wheel comes with a cost. Wheel is too powerful. Tested by centuries of time. Even vehicles that fly need wheels. Yet you might feel like trying and reimagining it. Here is my advice — just know when . Just know when you can afford to be innovative, and when it’s a death sentence (wheel). Anytime you feel the urge, just ask yourself “Is this a good time, or should I just force myself to do it as everyone expects me to?”. Trust me, you will get better at answering this question over time. just know when Potential employers don’t trust you After I finished studying design, I was looking for a job either as a UX/UI designer or a frontend developer. I ended up working as a B2B marketing expert. Yeah.. Even though I can code — just launched an app — and have been practicing UX/UI design for more than 8 years, employers generally don’t trust me when they see all these skills in my CV. This is something us generalists simply have to reckon with. app app I recommend making multiple CVs — one for each of your skills. Also, we will have a much harder time building a strong portfolio in multiple areas, compared to a specialist, who is laser focused on a single skill. Again, it will just take more time for us.. Thrive on chaos Trying to get your life perfectly sorted out might be great for most people. Not for most generalists. But don’t worry about it. Or try not to. Continue exploring. I know it’s hard with all the looks and pressure from your family and peers who try hard to follow linear trajectories. Startups. Your perfect environment? Startups are companies that are just starting out. While their resources might not be scarce, they need to spend them wisely. If they can hire a lawyer and a data scientist in one person (I met this guy!), they will gladly do it, while a serious law firm might be like “What?!”. Also, there are not as many rules at a startup. The company is new, responsibilities are not strictly divided, rigorous processes are not yet in place. As a generalist, you probably enjoy wearing multiple hats, and startups reward you for wearing them! Don’t be afraid As a generalist, fear is your greatest enemy . People fear the unknown, but the unknown and the generalist are the very best friends. Don’t be afraid to accept tasks or jump into things you have zero experience with. This is where your adaptability shines. Jump in, adapt, figure out. You will likely have a much easier time doing this than other people would. fear is your greatest enemy Being broke If you are broke, the natural tendency is to find a job. With too many skills, this might prove surprisingly difficult. If you struggle, use your creativity. Try to come up with a business idea. Chances are your ideas will be above average. Use your wide range of skills to offer a unique service. I know this advice is not much. You will probably have to find the job eventually. Just know that thinking outside the box is okay even when it comes to money. Get ready to burn out I don’t know. I still struggle with this one. I guess just get ready for it. You juggle 10 balls cause you have to. Juggling 3 doesn’t quite stimulate you, you will end up bored and you will procrastinate. Too much = burnout, too little = boredom and procrastination. If anyone has tips on how to find a balance here, please let me know. To sum it up: To sum it up: Your strengths: Your strengths: Extremely adaptable Quickly acquire new skills Eat challenges Extremely adaptable Extremely adaptable Quickly acquire new skills Quickly acquire new skills Eat challenges Eat challenges Your weaknesses: Your weaknesses: Huge time sacrifice Identity struggles Burnout Huge time sacrifice Huge time sacrifice Identity struggles Identity struggles Burnout Burnout Key takeaways: Key takeaways: Don’t resist, embrace Prepare for the long run Know when to be innovative Don’t fear the challenge Try balance burnout and boredom Don’t resist, embrace Prepare for the long run Know when to be innovative Don’t fear the challenge Try balance burnout and boredom

How To Survive as a Generalist

Imagine engaging in a thought-provoking conversation with an author, delving deep into the core ideas and exploring nuances often overlooked in summaries. BookNote.AI turns this scenario into reality, allowing you to grasp the essence of a book and decide if it’s worth your time. The Genesis of BookNote.AI At WebLab Technology, our team of AI-focused engineers is passionate about pushing the boundaries in fields such as Conversational AI, AI Agents, Reinforcement Learning, and Generative NLP. Beyond our tech pursuits, many of us are avid readers, ingrained with a love for books that shapes our daily routines. Our company’s “unlimited books” policy, affectionately known as “#bookworms,” allows every team member access to an extensive range of books, irrespective of their relevance to our work and available in any language. This policy has helped us build an impressive and diverse library, reflecting our belief that knowledge should have no boundaries. Amidst this culture of reading and innovation, the idea for BookNote.AI was conceived: What if we could harness our AI expertise to optimize the book discovery process? What if we could create a tool that not only helps find exciting reads but also facilitates deep, meaningful discussions about them? Thus, BookNote.AI was born — a platform designed to revolutionize how we interact with books, enabling a deeper connection with literature. What Sets BookNote.AI Apart? The traditional approach of reading books cover to cover can be daunting these days. Time is scarce whether you’re looking for an alternative to video reviews or tackling an ever-growing reading list. That’s where quick book summaries come in handy, distilling key ideas into concise overviews. They solve several critical pain points: Time Constraints: Modern life demands often leave little room for extensive reading. Summaries provide high-level insight without dedicating hours to reading entire books.
Information Overload: With an overwhelming number of books, it’s tough to choose which are truly worthwhile. Concise overviews help readers make informed selections.
Retention Issues: Even when books are read cover-to-cover, retaining every key point is difficult. Summaries serve as a refresher for understanding.
Optimal Decision-Making: For personal growth or professional development, grasping core book concepts can lead to better decisions and time saved. While services like GetAbstract, Blinkist, Instaread, Storyshots, and MakeHeadway condense book content, BookNote.AI takes it further by introducing an interactive, personalized reading experience. This allows deeper engagement where books can showcase their value firsthand — a unique approach other services can’t match. At its core, BookNote.AI provides the ability to quickly uncover profound insights from any book with the help of an AI assistant skilled in deepening and elevating discussions. For titles that cannot be legally indexed, we have prepared manually curated lists of booknotes and questions. The process is intuitive: Search and Select: Easily find your desired book on our intuitive search page.
Engage and Explore: Dive into interactive discussions about the book with our AI assistant.
Learn and Leap: Quickly grasp key ideas, improving your understanding and saving time. Interactivity and Personalization: A New Approach to Engaging with Books As we’ve mentioned before, BookNote.AI is designed to transform how you interact with books, offering a personalized and immersive experience that transcends simple summaries. Powered by AI, this platform enables you to engage in a dynamic dialogue with the book, allowing you to discover its core themes, ideas, and references. Let the book interest you through interactive communication Unlike static annotations, our platform supports specific questions and two-way dialogue, leading to a deeper understanding of a book’s content and themes. It’s similar to having a conversation with the author themselves. You are placed within the context of a particular book, allowing you to explore it as thoroughly as you need. At the same time, **BookNote.AI **makes it easier to share your reading journey with others. Often, we want to discuss what we’ve read or what we’ve learned, and we wish we could capture every thought. Unfortunately, it’s nearly impossible to convey every detail. Our team has improved this by allowing users to share their interactions with the AI, highlight specific moments, and show how their conversations about the book progressed. A peek behind the curtain: How does BookNote.AI work? At the heart of BookNote.AI is an AI assistant, carefully trained to read and analyze books. This assistant is the centerpiece of our platform, possessing a remarkable ability to analyze and interpret text. When you start chatting with the AI about a book, our LLM-based AI assistant becomes your guide, using advanced natural language processing and text summarization techniques. It explains the main ideas and themes of the text and can adjust its responses to any requested style. For example, if you ask it to explain an idea from a book as if to a 5-year-old, it will present the answer in that exact format. You transition from being a passive observer to an active participant in the dialogue with the book. You determine how the AI should present the book’s context to you and in what way. Rest assured, we plan to reveal the workings of our solution in upcoming articles. We will focus on examining specific components, highlighting their functions and details. Additionally, we may consider making our solution open-source when we believe it is ready to be shared with the community. WebLab Technology, a leading B2B SaaS company, boasts over 10 years of experience in assembling dedicated teams for complex projects. Led by Oleksandr Knyga, our team started active work with AI in 2019, tackling classical heuristic optimization problems and computer vision projects. Specializing in technologies such as AI, Cloud, and SaaS, we have established a reputation for delivering innovative and efficient solutions for our clients, including Shaman, AstraZeneca, Bryj, and Emaar Properties. Imagine engaging in a thought-provoking conversation with an author, delving deep into the core ideas and exploring nuances often overlooked in summaries. BookNote.AI turns this scenario into reality, allowing you to grasp the essence of a book and decide if it’s worth your time. BookNote.AI BookNote.AI The Genesis of BookNote.AI At WebLab Technology , our team of AI-focused engineers is passionate about pushing the boundaries in fields such as Conversational AI, AI Agents, Reinforcement Learning, and Generative NLP . Beyond our tech pursuits, many of us are avid readers, ingrained with a love for books that shapes our daily routines. WebLab Technology WebLab Technology Conversational AI, AI Agents, Reinforcement Learning, and Generative NLP Our company’s “unlimited books” policy, affectionately known as “ #bookworms, ” allows every team member access to an extensive range of books, irrespective of their relevance to our work and available in any language. This policy has helped us build an impressive and diverse library, reflecting our belief that knowledge should have no boundaries. #bookworms, #bookworms, Amidst this culture of reading and innovation, the idea for BookNote.AI was conceived: What if we could harness our AI expertise to optimize the book discovery process? What if we could create a tool that not only helps find exciting reads but also facilitates deep, meaningful discussions about them? BookNote.AI Thus, BookNote.AI was born — a platform designed to revolutionize how we interact with books, enabling a deeper connection with literature. BookNote.AI BookNote.AI What Sets BookNote.AI Apart? The traditional approach of reading books cover to cover can be daunting these days. Time is scarce whether you’re looking for an alternative to video reviews or tackling an ever-growing reading list. That’s where quick book summaries come in handy, distilling key ideas into concise overviews. They solve several critical pain points: Time Constraints: Modern life demands often leave little room for extensive reading. Summaries provide high-level insight without dedicating hours to reading entire books. Information Overload: With an overwhelming number of books, it’s tough to choose which are truly worthwhile. Concise overviews help readers make informed selections. Retention Issues: Even when books are read cover-to-cover, retaining every key point is difficult. Summaries serve as a refresher for understanding. Optimal Decision-Making: For personal growth or professional development, grasping core book concepts can lead to better decisions and time saved. Time Constraints: Modern life demands often leave little room for extensive reading. Summaries provide high-level insight without dedicating hours to reading entire books. Time Constraints: Information Overload: With an overwhelming number of books, it’s tough to choose which are truly worthwhile. Concise overviews help readers make informed selections. Information Overload: Retention Issues: Even when books are read cover-to-cover, retaining every key point is difficult. Summaries serve as a refresher for understanding. Retention Issues: Optimal Decision-Making: For personal growth or professional development, grasping core book concepts can lead to better decisions and time saved. Optimal Decision-Making: While services like GetAbstract, Blinkist, Instaread , Storyshots , and MakeHeadway condense book content, BookNote.AI takes it further by introducing an interactive, personalized reading experience. This allows deeper engagement where books can showcase their value firsthand — a unique approach other services can’t match. GetAbstract, GetAbstract, Blinkist, Blinkist, Instaread Instaread Storyshots Storyshots MakeHeadway MakeHeadway At its core, BookNote.AI provides the ability to quickly uncover profound insights from any book with the help of an AI assistant skilled in deepening and elevating discussions. For titles that cannot be legally indexed, we have prepared manually curated lists of booknotes and questions. BookNote.AI BookNote.AI The process is intuitive: Search and Select: Easily find your desired book on our intuitive search page. Engage and Explore: Dive into interactive discussions about the book with our AI assistant. Learn and Leap: Quickly grasp key ideas, improving your understanding and saving time. Search and Select: Easily find your desired book on our intuitive search page. Search and Select: Engage and Explore : Dive into interactive discussions about the book with our AI assistant. Engage and Explore Learn and Leap: Quickly grasp key ideas, improving your understanding and saving time. Learn and Leap: Interactivity and Personalization: A New Approach to Engaging with Books As we’ve mentioned before, BookNote.AI is designed to transform how you interact with books, offering a personalized and immersive experience that transcends simple summaries. Powered by AI, this platform enables you to engage in a dynamic dialogue with the book, allowing you to discover its core themes, ideas, and references. BookNote.AI BookNote.AI Let the book interest you through interactive communication Unlike static annotations, our platform supports specific questions and two-way dialogue, leading to a deeper understanding of a book’s content and themes. It’s similar to having a conversation with the author themselves. You are placed within the context of a particular book, allowing you to explore it as thoroughly as you need. At the same time, ** BookNote.AI **makes it easier to share your reading journey with others. Often, we want to discuss what we’ve read or what we’ve learned, and we wish we could capture every thought. Unfortunately, it’s nearly impossible to convey every detail. Our team has improved this by allowing users to share their interactions with the AI, highlight specific moments, and show how their conversations about the book progressed. BookNote.AI BookNote.AI A peek behind the curtain: How does BookNote.AI work? At the heart of BookNote.AI is an AI assistant, carefully trained to read and analyze books. This assistant is the centerpiece of our platform, possessing a remarkable ability to analyze and interpret text. BookNote.AI BookNote.AI When you start chatting with the AI about a book, our LLM-based AI assistant becomes your guide, using advanced natural language processing and text summarization techniques. It explains the main ideas and themes of the text and can adjust its responses to any requested style. For example, if you ask it to explain an idea from a book as if to a 5-year-old, it will present the answer in that exact format. You transition from being a passive observer to an active participant in the dialogue with the book. You determine how the AI should present the book’s context to you and in what way. Rest assured, we plan to reveal the workings of our solution in upcoming articles. We will focus on examining specific components, highlighting their functions and details. Additionally, we may consider making our solution open-source when we believe it is ready to be shared with the community. WebLab Technology , a leading B2B SaaS company, boasts over 10 years of experience in assembling dedicated teams for complex projects. Led by Oleksandr Knyga , our team started active work with AI in 2019, tackling classical heuristic optimization problems and computer vision projects. Specializing in technologies such as AI, Cloud, and SaaS, we have established a reputation for delivering innovative and efficient solutions for our clients, including Shaman , AstraZeneca , Bryj , and Emaar Properties. WebLab Technology WebLab Technology WebLab Technology Oleksandr Knyga Oleksandr Knyga AI, Cloud, and SaaS, Shaman Shaman AstraZeneca AstraZeneca Bryj Bryj Emaar Properties. Emaar Properties.

Transforming the Reading Experience with BookNote.AI by WebLab Technology

WebLab প্রযুক্তির দ্বারা BookNote.AI-এর মাধ্যমে পড়ার অভিজ্ঞতার রূপান্তর

Das Leseerlebnis verändern mit BookNote.AI von WebLab Technology

Transformando la experiencia de lectura con BookNote.AI de WebLab Technology

Transformer l'expérience de lecture avec BookNote.AI par WebLab Technology

वेबलैब टेक्नोलॉजी द्वारा BookNote.AI के साथ पठन अनुभव में परिवर्तन

WebLab Technology の BookNote.AI で読書体験を変革

WebLab 기술을 통해 BookNote.AI로 독서 경험을 혁신하다

Transformando a experiência de leitura com BookNote.AI da WebLab Technology

Преобразование опыта чтения с помощью BookNote.AI от WebLab Technology

WebLab Technology'nin BookNote.AI ile Okuma Deneyimini Dönüştürüyor

Chuyển đổi trải nghiệm đọc với BookNote.AI của WebLab Technology

WebLab Technology 推出 BookNote.AI 改变阅读体验

The Press Release is NOT Dead (Neither is the Funding Announcement)

প্রেস রিলিজ মৃত নয় (ফান্ডিং ঘোষণাও নয়)

Die Pressemitteilung ist NICHT tot (ebenso wenig wie die Finanzierungsankündigung)

El comunicado de prensa NO está muerto (tampoco el anuncio de financiación)

Le communiqué de presse n’est PAS mort (l’annonce de financement non plus)

प्रेस विज्ञप्ति ख़त्म नहीं हुई है (न ही फंडिंग की घोषणा)

プレスリリースは死んでいない（資金調達の発表も同様）

보도 자료는 죽지 않았습니다 (펀딩 발표도 마찬가지입니다)

O comunicado de imprensa NÃO está morto (nem o anúncio de financiamento)

Пресс-релиз НЕ умер (как и объявление о финансировании)

Basın Bülteni ÖLÜ DEĞİLDİR (Finansman Duyurusu da Değil)

Thông cáo báo chí KHÔNG chết (Thông báo tài trợ cũng không)

新闻稿还没有过时（融资公告也没有过时）

How to Use Knowledge Graphs for Retrieval-Augmented Generation—Without a Graph DB

পুনরুদ্ধার-অগমেন্টেড জেনারেশনের জন্য জ্ঞান গ্রাফগুলি কীভাবে ব্যবহার করবেন—একটি গ্রাফ ডিবি ছাড়াই

So nutzen Sie Wissensgraphen zur abruferweiterten Generierung – ohne Graph-DB

Cómo utilizar gráficos de conocimiento para la generación de recuperación aumentada, sin una base de datos de gráficos

Comment utiliser les Knowledge Graphs pour la génération augmentée par récupération – sans base de données graphique

रिट्रीवल-ऑगमेंटेड जेनरेशन के लिए नॉलेज ग्राफ़ का उपयोग कैसे करें - बिना ग्राफ़ DB के

グラフ DB を使用せずにナレッジ グラフを検索拡張生成に使用する方法

검색 증강 생성을 위해 지식 그래프를 사용하는 방법 - 그래프 DB 없이

Como usar gráficos de conhecimento para geração aumentada de recuperação - sem um banco de dados gráfico

Как использовать графы знаний для генерации с расширенным поиском — без базы данных графов

Almayla Artırılmış Oluşturma için Bilgi Grafikleri Nasıl Kullanılır - Grafik Veritabanı Olmadan

Cách sử dụng Sơ đồ tri thức cho thế hệ tăng cường truy xuất—Không cần cơ sở dữ liệu đồ thị

如何使用知识图谱进行检索增强生成（无需图形数据库）

Retrieval augmented generation (RAG) is by now of generative artificial intelligence (AI) applications. Supplementing your application prompt with relevant context retrieved from a vector database can dramatically increase accuracy and reduce hallucinations. This means that increasing relevance in vector search results has a direct correlation to the quality of your RAG application. a standard part There are two reasons RAG remains popular and increasingly relevant even : as large language models (LLMs) increase their context window LLM response time and price both increase linearly with context length. LLMs still struggle with both and across massive contexts. retrieval reasoning But isn’t a magic wand. In particular, the most common design, dense passage retrieval (DPR), represents both queries and passages as a single embedding vector and uses straightforward to score relevance. This means DPR relies heavily on the embeddings model having the breadth of training to recognize all the relevant search terms. RAG cosine similarity Unfortunately, off-the-shelf models struggle with unusual terms, including names, that are not commonly in their training data. DPR also tends to be hypersensitive to chunking strategy, which can cause a relevant passage to be missed if it’s surrounded by a lot of irrelevant information. All of this creates a burden on the application developer to “get it right the first time,” because a mistake usually results in the need to rebuild the index from scratch. Solving DPR’s challenges with ColBERT is a new way of scoring passage relevance using a language model that substantially solves the problems with DPR. This diagram from shows why it’s so exciting: ColBERT BERT the first ColBERT paper This compares the performance of ColBERT with other state-of-the-art solutions for the MS-MARCO dataset. (MS-MARCO is a set of Bing queries for which scored the most relevant passages by hand. It’s one of the better retrieval benchmarks.) Lower and to the right is better. Microsoft In short, ColBERT handily outperforms the field of mostly significantly more complex solutions at the cost of a small increase in latency. To test this, I created and indexed over 1,000 Wikipedia articles with both ada002 DPR and ColBERT. I found that ColBERT delivers significantly better results on unusual search terms. a demo The following screenshot shows that DPR fails to recognize the unusual name of William H. Herndon, an associate of Abraham Lincoln, while ColBERT finds the reference in the Springfield article. Also note that ColBERT’s No. 2 result is for a different William, while none of DPR’s results are relevant. ColBERT is often described in dense machine learning jargon, but it’s actually very straightforward. I’ll show how to implement ColBERT retrieval and scoring on with only a few lines of Python and (CQL). DataStax Astra DB Cassandra Query Language The big idea Instead of traditional, single-vector-based DPR that turns passages into a single “embedding” vector, ColBERT generates a contextually influenced for each token in the passages. ColBERT similarly generates vectors for each token in the query. vector (Tokenization refers to breaking up input into fractions of words before processing by an LLM. .) Andrej Karpathy, a founding member of the OpenAI team, just released an outstanding video on how this works Then, the score of each document is the sum of the maximum similarity of each query embedding to any of the document embeddings: def maxsim(qv, document_embeddings):
    return max(qv @ dv for dv in document_embeddings)

def score(query_embeddings, document_embeddings):
    return sum(maxsim(qv, document_embeddings) for qv in query_embeddings) (@ is the PyTorch operator for dot product and is the most common .) measure of vector similarity That’s it — you can implement ColBERT scoring in four lines of Python! Now you understand ColBERT better than 99% of the people posting about it on X (formerly known as Twitter). The rest of the ColBERT papers deal with: How do you fine-tune the BERT model to generate the best embeddings for a given data set? How do you limit the set of documents for which you compute the (relatively expensive) score shown here? The first question is optional and out of scope for this writeup. I’ll use the pretrained ColBERT checkpoint. But the second is straightforward to do with a vector database like DataStax Astra DB. ColBERT on Astra DB There is a popular Python all-in-one library for ColBERT called ; however, it assumes a static dataset. One of the powerful features of RAG applications is responding to dynamically . So instead, I’m going to use Astra’s vector index to narrow the set of documents I need to score down to the best candidates for each subvector. RAGatouille changing data in real time There are two steps when adding ColBERT to a RAG application: ingestion and retrieval. Ingestion Because each document chunk will have multiple embeddings associated with it, I’ll need two tables: CREATE TABLE chunks (
    title text,
    part int,
    body text,
    PRIMARY KEY (title, part)
);

CREATE TABLE colbert_embeddings (
    title text,
    part int,
    embedding_id int,
    bert_embedding vector<float, 128>,
    PRIMARY KEY (title, part, embedding_id)
);

CREATE INDEX colbert_ann ON colbert_embeddings(bert_embedding)
  WITH OPTIONS = { 'similarity_function': 'DOT_PRODUCT' }; After installing the ColBERT library ( ) and downloading , I can load documents into these tables: pip install colbert-ai the pretrained BERT checkpoint from colbert.infra.config import ColBERTConfig
from colbert.modeling.checkpoint import Checkpoint
from colbert.indexing.collection_encoder import CollectionEncoder

from cassandra.concurrent import execute_concurrent_with_args
from db import DB


def encode_and_save(title, passages):
    db = DB()
    cf = ColBERTConfig(checkpoint='checkpoints/colbertv2.0')
    cp = Checkpoint(cf.checkpoint, colbert_config=cf)
    encoder = CollectionEncoder(cf, cp)

    # encode_passages returns a flat list of embeddings and a list of how many correspond to each passage
    embeddings_flat, counts = encoder.encode_passages(passages)

    # split up embeddings_flat into a nested list
    start_indices = [0] + list(itertools.accumulate(counts[:-1]))
    embeddings_by_part = [embeddings_flat[start:start+count] for start, count in zip(start_indices, counts)]

    # insert into the database
    for part, embeddings in enumerate(embeddings_by_part):
        execute_concurrent_with_args(db.session,
                                     db.insert_colbert_stmt,
                                     [(title, part, i, e) for i, e in enumerate(embeddings)]) (I like to encapsulate my DB logic in a dedicated module; you can access the full source in my .) GitHub repository Retrieval Then retrieval looks like this: def retrieve_colbert(query):
    db = DB()
    cf = ColBERTConfig(checkpoint='checkpoints/colbertv2.0')
    cp = Checkpoint(cf.checkpoint, colbert_config=cf)
    encode = lambda q: cp.queryFromText([q])[0]

    query_encodings = encode(query)
    # find the most relevant documents for each query embedding. using a set
    # handles duplicates so we don't retrieve the same one more than once
    docparts = set()
    for qv in query_encodings:
        rows = db.session.execute(db.query_colbert_ann_stmt, [list(qv)])
        docparts.update((row.title, row.part) for row in rows)
    # retrieve these relevant documents and score each one
    scores = {}
    for title, part in docparts:
        rows = db.session.execute(db.query_colbert_parts_stmt, [title, part])
        embeddings_for_part = [tensor(row.bert_embedding) for row in rows]
        scores[(title, part)] = score(query_encodings, embeddings_for_part)
    # return the source chunk for the top 5
    return sorted(scores, key=scores.get, reverse=True)[:5] Here is the query being executed for the most-relevant-documents part ( ): db.query_colbert_ann_stmt SELECT title, part
FROM colbert_embeddings
ORDER BY bert_embedding ANN OF ?
LIMIT 5 Beyond the basics: RAGStack This article and the linked repository briefly introduce how ColBERT works. You can implement this today with your own data and see immediate results. As with everything in AI, best practices are changing daily, and new techniques are constantly emerging. To make keeping up with the state of the art easier, DataStax is rolling this and other enhancements into , our production-ready RAG library leveraging LangChain and LlamaIndex. Our goal is to provide developers with a consistent library for RAG applications that puts them in control of the step-up to new functionality. Instead of having to keep up with the myriad changes in techniques and libraries, you have a single stream, so you can focus on building your application. You can use RAGStack today to incorporate best practices for LangChain and LlamaIndex out of the box; advances like ColBERT will come to RAGstack in upcoming releases. RAGStack By Jonathan Ellis, DataStax Also appears . here

How ColBERT Helps Developers Overcome the Limits of RAG

কিভাবে ColBERT ডেভেলপারদের RAG এর সীমা অতিক্রম করতে সাহায্য করে

Wie ColBERT Entwicklern hilft, die Grenzen von RAG zu überwinden

Cómo ColBERT ayuda a los desarrolladores a superar los límites de RAG

Comment ColBERT aide les développeurs à surmonter les limites de RAG

कैसे ColBERT डेवलपर्स को RAG की सीमाओं को पार करने में मदद करता है

ColBERT が開発者による RAG の限界の克服にどのように役立つか

ColBERT가 개발자가 RAG의 한계를 극복하도록 돕는 방법

Como ColBERT ajuda os desenvolvedores a superar os limites do RAG

Как ColBERT помогает разработчикам преодолеть ограничения RAG

ColBERT, Geliştiricilerin RAG Sınırlarını Aşmasına Nasıl Yardımcı Olur?

ColBERT giúp nhà phát triển vượt qua giới hạn của RAG như thế nào

ColBERT 如何帮助开发人员克服 RAG 的限制

Recommendation systems have become an integral and indispensable part of our lives. These intelligent algorithms are pivotal in shaping our online experiences, influencing the content we consume, the products we buy, and the services we explore. Whether we are streaming content on platforms like , discovering new music on , or shopping online, recommendation systems are quietly working behind the scenes to personalize and enhance our interactions. Netflix Spotify The unique element of these recommendation systems is their ability to understand and predict our preferences based on historical behavior and user patterns. By analyzing our past choices, these systems curate tailored suggestions, saving us time and effort while introducing us to content/products that align with our interests. This enhances user satisfaction and fosters discovery, introducing us to new and relevant offerings that we might not have encountered otherwise. At a high level, developers understand that these algorithms are powered by machine learning and deep learning systems (interchangeably called neural networks), but what if I tell you there is a way to build a recommendation engine without going through the pain of deploying your neural net or machine learning model? This question is specifically relevant in the context of early and mid-stage startups because they don't have tons of structured data to train their models. And as we already know, most machine learning models will not give accurate predictions without proper training data. I recently built and deployed a basic recommendation engine for a , which led to a 40% jump in our key metrics. At the time of writing this blog, the system is generating more than 30 million recommendations per month. Even though this recommendation system was built for a social network, you can apply the basic architecture to any use case, such as product recommendations, music recommendations, content recommendations on text and video platforms, or anything else. Let me start by describing the problem statement. voice-first social network Problem statement from the engineering perspective I had an extensive and because we were building the recommendation system for a product that is already used by thousands of users daily. But to keep this blog short and on point, I will write only the high-level requirements and then discuss the solution of the same. If you are building a recommendation system for your product (simple or neural net-based) and are stuck somewhere, please feel free to contact me on or and I will be more than happy to answer your questions. product requirement document subsequent engineering requirements document Twitter , Linkedin At a high level, we had the following requirements from an engineering perspective - The system should be able to capture a user's interests in the form of keywords. The system should also be able to classify the level of interest a user has with specific keywords. The system should be able to capture a user's interest in other users. It should be able to classify the level of interest a user has in content created by another user. The system should be able to generate high-quality recommendations based on a user's interests. The system should be able to ensure that the recommendations already viewed/rejected by the user shouldn't re-appear again for X number of days. The to ensure that the posts from the same creators aren't grouped on the same page. The system should try its best to ensure that if a user consumes ten posts (our page size), all of those should be from different creators. system should have logic The system should be fast. Less than 150 milliseconds of P99 latency. All the other non-functional requirements, such as high availability, scalability, security, reliability, maintainability, etc, should be fulfilled. Again, this is a highly oversimplified list of problem statements. In reality, the documents were 3000+ words long as they also covered a lot of edge cases and corner cases that can arise while integrating this recommendation engine into our existing systems. Let's move on to the solution. Solution - High-level working of the recommendation engine I will discuss the solutions to the problem one by one and then will describe the overall working of the entire system. Our first problem is capturing the user's interests and defining their interest level with a specific interest For this, we created something called a . To put it simply, a social graph stores the relationships and connections between different entities in a social network. These entities can be different users or a relationship of users with a specific interest. Social graphs are a powerful way to understand and structure the relationships within a particular system. For the sake of brevity, I will not explain the social graph in detail, but I will recommend you google it and learn more about it. Following is a simplified version of the social graph I built for our recommendation engine. social graph As you can see from the above image, we are storing a lot of information, such as the number of interactions (likes, comments, shares) and recency of these interactions (when they happened last) as relationship data between two users as well between a user and an interest. We are even storing the relationship between two different interest keywords. I used , a managed graph database by AWS, to store this social graph. You can use any other graph database, such as , etc. Amazon Neptune Neo4j, JanusGraph, ArrangoDB These interest keywords are predominantly nouns. There is a system in place that breaks down the contents of a post into these keywords(nouns). It's powered by a natural-language processing (NLP) service that uses machine learning to break text into entities, key phrases, etc. Again, you can use any managed NLP services (several available) to accomplish the same. You don't need to learn or deploy your machine-learning models! If you already understand machine learning, then you can go check . AWS comprehend; open-source NLP models as well on Huggingface Our second problem is generating high-quality recommendations based on a user's interest The following diagram is a simplified high-level representation of how the system works. While the above looks easy, there is a lot more going on at each step, and those things have to be carefully thought through and then programmed to ensure that the system is performing optimally. Let me explain step by step: Step 1 - Converting post content into vector embeddings To generate these recommendations, first, we have to convert the contents of a post into something called - . With the recent uptick in the branding of LLMs, OpenAI( the makers of ChatGPT) and , Vector embeddings are becoming an everyday term. I will not go into the details of what they are and how they work, but I highly recommend reading more about them. But generating viable candidates for a feed also has to account for things like content privacy and moderation (removing profane words, abuses, sexual content, harassment, filtering blocked users, etc). Vector embeddings Vector databases For generating the vector embeddings, you can use any prominent embedding model such as the , or any text embedding model, depending on your use case. We went with because of its friendly pricing, performance and operational ease. OpenAI embedding model Amazon titan open-source Amazon Titan Step 2 - Query the user's interest Now, this is where things get interesting. You would want to design the queries based on your specific business needs. For example, we give more weightage to the recency of engagement while querying interests than the number of engagements with a specific keyword or user. We also run multiple parallel queries to find different types of interest of the user - keyword or other user. Since we generate multiple feeds for a single user, we also run some queries promoting a specific topic according to the trend (for example, you will see many Christmas-related posts near Christmas or earthquake-related posts if some earthquake has happened). Needless to say, this topic will only come up in the query results if the user has expressed some interest in them in their journey. So, choose the logic that suits your business use case and the behavior that you want to drive and run multiple queries to get a big enough list of all the user's interests. Step 3 - Do an ANN search based on the interests found Vector databases are predominantly used for performing a particular type of search called (ANN). Again, the way you categorize various interests and whether you are doing one big ANN search or parallel difference searches should entirely be based on your use case and business requirements. I recommend doing more than cohort-based searches and then ordering the results (we will discuss this later in this blog) for the best end-user experience. What ANN search does, in this case, is find other posts on the platform that are similar (closer) to the interests of the user. Approximate nearest neighbor search Step 4 - Store the results in a cache database with ordering. Cache database because one of the problems that we need to solve is speed. We used for storing the unique IDs of the posts for a specific user. We used redis sorted sets because the order of posts in a user's feed is critical. Also, another problem that you have to solve is that the" to ensure that the posts from the same creators aren't grouped on the same page". To avoid repetition of content from the same creator, we have written a simple algorithm which ensures that if a specific creator's post is inserted at any position in a particular user's feed (sorted set), we don't insert another post from the same creator for successive ten positions (we have a page size of 10 while serving the feed to the end user, so we kept it static to avoid complexity). sorted sets redis system should have logic For deciding the order of a specific recommendation of the user, we factored in the following things - : It's determined by an arithmetic formula that takes various data points from the social graph. All of this is engagement data like the timestamp of the last likes created, number of likes created, last comment, etc. User engagement behavior is the indicator of their interest in something. The strength of the relationship with a specific interest (or another user) for this user To determine this, we have created an algorithm that takes various factors such as engagement, engagement-to-impression ratios, number of unique users who engaged, etc., to generate an engagement score of that post at a platform level. The popularity of the post on the platform: In some feeds, we prioritize popularity; in others, we prioritize the social graph. But mostly, all of them are a healthy mix of the two. How the System Works As you can see from the diagram above, the system has been intentionally kept very simple. Following is how the system works - When user A creates a post, the post service, after saving that post, triggers a pub/sub event to a queue, which is received by a background service meant for candidate generation. We use for the pub/sub functionality. Google Pub/Sub This background service receives this asynchronously and performs functionalities discussed earlier - Privacy checks, moderation checks, and keyword generation and then generates the vector embeddings and stores them in the vector database. We are using (discussed later). AstraDB as our vector database Whenever a user engages (like/comment/share, etc.) after updating our main NoSQL database, the post-service triggers a pub/sub event to the recommendation engine service. This recommendation engine service updates the graph database and then updates the recommended feed of the user in near real-time by performing the ANN search and updating the Redis database. So, the more users interact, the better the feed keeps getting. . Those checks are performed while we query the Graph database. This service also updates the engagement score asynchronously. Engagement scores are re-calculated on users viewing the post as well. There are checks to ensure that the recommendations are not biased towards a specific list of keywords Since all of the above steps are performed asynchronously behind the scenes, these computations have no impact on the end-user experience. The feed is finally served to the end user through a feed service. Since this service just performs a lookup on redis and our main NoSQL database ( ), its P99 latency is less than 110 milliseconds. Both these databases return query results in single-digit millisecond latency irrespective of scale. DyanmoDB Tools and technologies used Some services have been written in , while others have been written in (with typescript). Go programming language NodeJS We are using as our vector database. We arrived at this decision after evaluating multiple other databases, such as and . Apart from its excellent query and indexing capabilities on vector and other data types, it offers a pocket-friendly serverless pricing plan. It runs on top of a Cassandra engine, which we use as a database in several other features on our platform, and it gives a CQL query interface, which is very developer-friendly. I highly recommend trying it for your vector use cases. AstraDB by Datastax pinecone, milvus weaviate We use for our asynchronous communication because, at our current scale (few lakh total users, few thousand daily active users), it's highly cost-effective. I have run it at a scale of a few lakh users with thousands of events per second. It works well, and it's effortless to use and extend. Google pub/sub - Speed, simplicity and powerful data structure. I don't think I need to discuss why redis in 2024. Redis - Again, it is highly scalable and easy to use, and we run it in the serverless mode where, despite hundreds of thousands of queries per minute, our total bill is quite low. It also offers very powerful indexing capabilities and single-digit millisecond latency in reads and writes. DynamoDB Problems to be solved in the future As you can imagine, this same setup can be tweaked to build a basic recommendation engine for any use case. But, since ours is a social network, we will require some tweaks down the line to make this system more efficient. Machine learning/ Deep learning algorithms will be needed at the social graph level to predict the keywords and users most relevant for the user. Currently, the data set is too small to predict anything accurately as it is a very new product. However, as the data grows, we will need to replace the current simple queries and formulas with the output of machine learning algorithms. Relationships between various keywords and users must be fine-tuned and made more granular. They are at a very high level right now. But they will need to be deeper. We will need to explore the second and third-degree relationships in our graph to refine the recommendations first. We are not doing any fine-tuning in our embedding models right now. We will need to do that in the near future. End note I hope you found this blog helpful. If you have any questions, doubts or suggestions, please feel free to contact me on , or . Do share this article with your friends and colleagues. Twitter Linkedin Instagram Also published . here

How to Build a Basic Recommendation Engine without Machine Learning 

কিভাবে মেশিন লার্নিং ছাড়া একটি বেসিক রেকমেন্ডেশন ইঞ্জিন তৈরি করবেন

So erstellen Sie eine einfache Empfehlungs-Engine ohne maschinelles Lernen

Cómo construir un motor de recomendación básico sin aprendizaje automático

Comment créer un moteur de recommandation de base sans apprentissage automatique

मशीन लर्निंग के बिना एक बुनियादी अनुशंसा इंजन कैसे बनाएं

機械学習を使用せずに基本的なレコメンデーション エンジンを構築する方法

머신러닝 없이 기본 추천 엔진을 구축하는 방법

Como construir um mecanismo de recomendação básico sem aprendizado de máquina

Как создать базовую систему рекомендаций без машинного обучения

Makine Öğrenimi Olmadan Temel Öneri Motoru Nasıl Oluşturulur

Cách xây dựng công cụ đề xuất cơ bản mà không cần học máy

如何在不使用机器学习的情况下构建基本的推荐引擎

So let’s start! Tell us a bit about yourself. For example, name, profession, and personal interests. My name is Dmitrii (for most non-Russian speaking people, it is easier to pronounce as ‘Dimitry’). I am a software developer since the 1990s when I first started with Borland tools, such as Paradox and Delphi. In the mid-2000s, I switched to .NET, and my first project for it was developed in version 1.1. Quickly after version 2.0 was released, the project was updated to the new .NET version, and since then, I have been doing all sorts of applications - WinForms and WPF, Web APIs, Silverlight, WCF services, cloud-hosted apps and services, and on and on... I like to emphasize on code quality and clean code (not the same as the 'clean architecture', mind you!) - this is what I always keep in mind while designing, implementing, or doing code reviews. And do not take me wrong - I have nothing against clean architecture; it is just a very different level of concern that has its own time and place. Interesting! What was your latest Hackernoon Top story about? The is about log enrichment; it is my second story dedicated to improving the code quality around logging and observability. Unfortunately, my personal observation is that many projects suffer from treating these concerns as 'second-class citizens', often neglecting best practices and patterns when it comes to adding or implementing logging and observability. latest story Thus, since one of my recent efforts in a large-scale enterprise system was closely related to cleaning up exactly this problem, I decided to share with a wider audience what can done to lessen the pain and improve the quality. Do you usually write on similar topics? If not, what do you usually write about? Actually, I do not write that often, not until recently, at least. Mostly, it is live communication with other people in my company over a topic. However, as I realized, most of these interactions revolve around my favorite subjects of code quality, and this is what I would like to explore further in writing. Great! What is your usual writing routine like (if you have one?) Not much routine here, yet. However, it all begins with live code, and then I can see how some particular code piece or approach can be extended to other similar cases, generalized, and made useful for other people and situations. So, I take this code piece, clean it up - anonymize, if you will - and generalize to better suit the purpose of demonstrating a particular concept. Then I go over the code and explore all important bits and pieces - 'the devil is in the details' - generally it is not enough to copy and paste a piece of code, or call a method, or use a class. A developer needs to know the why's and side effects - there may be a real game-changer that may turn a proposed solution into its opposite if these details are not respected. Being a writer in tech can be a challenge. It’s not often our main role, but an addition to another one. What is the biggest challenge you have when it comes to writing? I think the most challenging is staying brief and simple. Too many words obscure the meaning and may lead a reader astray. The writer is also in danger of not reaching the initial goal while probably not even realizing it. On a humorous note - the 'single responsibility principle' shines in many aspects of our life; not just in software development. What is the next thing you hope to achieve in your career? Having more impact - taking on a role that would allow me to help more and be involved in doing what matters to many people and businesses. Wow, that’s admirable. Now, something more casual: What is your guilty pleasure of choice? That would be good wine and liquor... In measure, of course. I enjoy exploring wines and strong drinks offered by different countries and cultures in the company of my friends. Good and fitting wine may change a regular meal into something really memorable and enjoyable. Elegance really affects the quality - be it in engineering or in more simple life experiences. Do you have a non-tech-related hobby? If yes, what is it? Traveling. I did not visit that many places but I can count stamps of probably fifteen countries in my passport, and I do not mind coming back again and exploring more. The world is very rich, we only need to take what it offers. What can the Hacker Noon community expect to read from you next? I think more on code quality in that or another way. As I mentioned, the main challenge I see is in being brief and simple. Quality often comes with simplicity (not via simplification, though). But making things simple is often one of the hardest things to do, thus there is a lot to learn and share here. What’s your opinion on HackerNoon as a platform for writers? Actually, I am grateful - Hackernoon gives a convenient way of sharing with other people, and it’s effective and far-reaching. A real treat for both a reader and a writer. Thanks for taking time to join our “ ” series. It was a pleasure. Do you have any closing words? Meet the writer Thank you for the opportunity to share!

Meet the Writer: HackerNoon Contributor Dmitrii - A Software Developer Since the 90s

When building a generative AI application that needs to call a large language model (LLM) multiple times to complete a task, a common problem is that repeated queries to the LLM can be both expensive and unpredictable. Large models such as GPT-3.5/4 are incredibly resource-intensive to train and run inference; this is reflected in the API charges as well as occasional disruptions to service. ChatGPT was originally released as a research preview and wasn’t intended to be used for production applications. However, its usefulness across a vast array of applications is indisputable, so has exploded. interest in LLMs Since the inception of ChatGPT, users have been looking for ways around the lack of privacy and the inability to control uptime or inference settings when using GPT. This has led to the popularity of free, public models such as Meta’s and later the creation of quantized and lower-parameter versions of Llama that can run on consumer hardware. These public models are capable of providing much of the same functionality as GPT for much less computing power, though at the cost of fewer parameters and fewer verbose outputs. Llama 2 If your application doesn’t necessarily depend on processing excessively large contexts or producing verbose outputs, then hosting your own inference on instances you control can be a more cost-effective option. And when it comes to real-world applications of (RAG), the cost differences may become even more significant. retrieval augmented generation I’ll demonstrate a simple method for combining vector stores, lexical search, and prompt engineering to perform accurate RAG on commodity hardware. Using this method, you can both reduce the complexity of large volumes of information and make running generative AI applications more accurate, efficient, and cost-effective at scale. By using RAG on specific stores of information, you can gain the ability to eliminate hallucinations and create effective and knowledgeable agents from any source material without needing to pay for third-party APIs. To get started, you will need either a DataStax Enterprise 7 instance or to store the vectors and text data, as well as an LLM and a sentence transformer model to generate responses and encode your data with vectors. Depending on the complexity of your data or the user prompts, you may also consider combining this with a DataStax Enterprise 6.8 database that can perform searches to match wider ranges of data, which is what I used in this example. DataStax is continually working on improvements to enable all of these operations with a single database, but for now, I use two databases. DataStax Astra DB Solr Solving for Hallucinations Regardless of which LLM you choose, they all still suffer from . For now, that limitation needs to be resolved by feeding truthful information into the context of prompts to the LLM, otherwise known as RAG. The method by which you locate your information and transform it for the prompts completely depends on your data model, but you can find more pertinent information in a more efficient way by using vector databases. hallucinations Say, for instance, that you have a collection of e-books on a subject you’d like to explore, such as how to play . Under normal circumstances, it would take years to read through the supporting literature and gain enough gameplay experience to reach an expert level. Warhammer 40,000 A targeted question such as, “What can you tell me about Morvenn Vahl of Adepta Sororitas?” would be best answered by a veteran player or any employee at a Warhammer store. While ChatGPT can answer many questions about the game, it unfortunately has no training data that covers this particular character: Compare this to a Llama 2 13B parameter LLM hosted on a consumer workstation with an Nvidia RTX A4000 graphics card. Similarly, the model can demonstrate basic knowledge of the Warhammer universe, but because of the tuning, the model doesn’t care that the character isn’t found and provides instead a best-effort hallucination: If you want to build a chatbot that can help both newcomers and veterans play Warhammer 40,000, then these outputs are unacceptable. To be an effective game guide, the chatbot needs to know the rules of the game, the rules for each unit, some bits of the lore, and some strategy and commentary. Luckily, all of that information on the 10th-edition rules is available for free from Games Workshop and fan websites, and all you need to do is make it searchable to your chatbot app. Compare this to the same 13B Llama model, where with RAG, it is asked to compare a couple of sources on Morvenn Vahl and devise a relevant answer based on the user prompt. This time, the chatbot has access to a database full of all the public information on how to play Warhammer 40,000, 10th Edition: search database and a vector What a difference! Not only does it find pertinent information on this niche character, but it also keeps its outputs in line with the context of how to play the game with the 10th-edition rules. The hardest part in all of this is performing an effective search to find the relevant pages to feed into the LLM. This is where vector databases can be particularly useful. Applying vectors In this example, we’ll use DSE 7 and DSE 6.8 running in Docker instances to satisfy the database requirements of the chatbot application, which needs to be able to compare vectors and perform lexical searches. DSE 7 and Astra DB have introduced the ability to store vectors and perform vector searches, as well as filtering by text matches. We only need to search a few dozen books for this example, so running DSE instances in Docker will be sufficient for most consumer hardware. Using vectors in your databases will help to find documents that are similar to a given query, or they can be used to compare results retrieved from another search. This can help you to overcome the limitations of lexical search and to improve the effectiveness of data models. For instance, something like e-book PDFs can benefit from being encoded with sentence transformers like , and the vectors can be used to run a similarity comparison between a query and a given source. In this case, a sentence transformer model is used to create embeddings of a page’s text in an e-book, and this can enable you to compare to the user’s prompt to figure out if a result is relevant to the query. Relevant pages should contain one or more instances of terms that are similar to the user’s query and result in better similarity scores from the model’s standpoint. miniLM That said, the vectors are best applied as a supplement to an existing lexical search model. If you search by vectors only, then you might end up unexpectedly retrieving unrelated documents and providing them as context where they don’t apply. In this example, the query “What can you tell me about Morvenn Vahl of Adepta Sororitas?” can be transformed by an LLM to a set of simple search terms: Morvenn, Vahl, Adepta, Sororitas The first step in finding relevant documents would be to search for documents that contain those basic terms. This can be done by first filtering for text matches in the database to find keywords in the page text matching such a query. The reason for using an LLM to generate keywords is to provide a wider range of possible keywords to search, as it often attempts to add more keywords that are related but are not in the text of the original prompt. Be careful with this, however, as LLMs can also generate special characters and odd sequences that you will need to sanitize. Once you have at least one result, you can vectorize the user’s query and compare it to the vectors of the lexical search, creating scores of how relevant each result is. This allows you to check the search results for accuracy to the query and set a threshold for rejecting unrelated results when it comes to finally presenting your results to the LLM. In this case, the first step should match pages that specifically show Morvenn Vahl’s index card or gameplay mechanics because those describe the character’s unit in terms of how it plays in the game. If the page meets a certain relevance threshold to the user query, determined by the application, then it gets summarized and placed in a list of results. Finally, the search results can be compiled into a list and fed back to the LLM, where it is asked to use the most relevant contexts to answer the original query. Here is a visualization of the flow: As you can see, the LLM gets called quite frequently for this flow. The LLM is responsible for transforming the user prompt into keywords, summarizing applicable results, and choosing which context best answers a query. Each source to check adds another LLM invocation, which can be quite expensive when making queries to GPT. But if you already have the information you need and just want to summarize it or transform it, then you may not need to use such a large model. In fact, switching to smaller models can provide a number of benefits. By using a smaller LLM, you can reduce the computational cost of each query, which can lead to significant savings over time. This can also result in faster response times for your users, which can improve their overall experience. In this example, where RAG is performed using a small LLM and small databases, all hosted on the same GPU instance, it takes about 30 seconds to retrieve 15 sources, analyze them for relevance, and provide a final answer. And the shorter the prompts (sources), the faster the outputs can be returned. Additionally, this method allows for increased security and scalability. With prompt engineering and a pipeline of calls to the LLM, you’re in is accessed and what the users will get in their responses. In terms of resource usage, the example 13B parameter model only consumes a little over 8GB of VRAM and still provides relevant answers. Depending on needs, this shows potential for even running RAG on myriad other platforms, such as user workstations and mobile devices. full control of how the data Controlling the output Prompt engineering is key to making RAG do exactly what you want. You are in control of how the chatbot interprets the data and the context under which it should be thinking. In this example, we want to ensure that the chatbot knows we are specifically seeking Warhammer information, so we can first ask it to help provide supporting context to the user’s query: Query: “<user query>” Give me a minimal, comma-separated list of Warhammer 40K keywords for a search engine. Respond with only the query. Do not use emojis or special characters. Answer: Warhammer 40,000 is full of terms and names that might appear in other, unrelated popular culture, so it is important to set the context of the RAG in the very first query. This context should be something available to select or modify if your application covers multiple contexts, such as if you need to cover multiple editions of the Warhammer game rules or combine them with the official lore books. Note that the user’s query is always encapsulated with quotes for this experiment. This helps the LLM to distinguish between the query that it’s attempting to directly answer and the separate prompt-engineering instructions, which it must not directly answer. The question/answer portion of the prompt can be adjusted to fit a particular context, but mostly, all you need to be able to do is inform the LLM what it should and should not directly respond to and how to respond. In this case, it’s safe to assume that the LLM has a general knowledge of the game universe since the series is reasonably popular and general information is available for free. The output of this first query helps generate some keywords to use in lexical search without us having to build a cheat sheet into our application. The lexical and vector comparisons can then be performed in the background, and a list of results is compiled for review by the LLM. Because the user’s original prompt is never directly answered with inference at the first step, the LLM only transforms what is found in a search and can easily be stopped from answering queries outside of its guard rails or knowledge base. If there are relevant results from the search: Query: “<user query>” Review these search results and use them to answer the query. Result 1 Result 2 etc. Answer: If there are no relevant results from the search: Query: “<user query>” Politely tell me you searched but could not find an answer for the query. Answer to the best of your knowledge instead. Answer: For added security, you can completely reject or redirect the request when it cannot be served. Query: “<user query>” Politely tell me you searched but could not find an answer to the query. Instruct me to reach out to the Customer Support team for assistance instead. Answer: You can even make the outputs lengthier by asking for more details. As long as you can fit your source material within the context window, the LLM can transform it for you. Query: “<user query>” Review these search results and use them to answer the query. Be as detailed as possible and cite the sources. Result 1 Result 2 etc. Answer: Limitations The LLM has a limited context window and will fail to process exceptionally large pages of text. Consider placing limits on row size so your and easier for the LLM to process. For example, cutting pages into chunks of about 1,000 characters seems to work well, and try to avoid feeding more than four or five detailed answers into the prompt. data is more manageable The LLM has no memory of a conversation aside from what you can fit in the context window. It is possible to build a permanent store of conversation data, but it is not possible for an LLM to fit excessively large conversations or detailed context into a prompt; there is an upper limit to what it can transform. This means that no matter what, at a certain point, you will notice that the LLM seems to “forget” certain details even when they are provided as context; this is just an inherent limitation of the tool. It is best to rely on it for short conversations only and focus on transforming small amounts of text at a time to minimize hallucinations. Randomness in the LLM can be a problem. Testing and tuning will be necessary to determine which prompts work best for your data set and to find which model works best for your use case. In my testing with a 13B parameter model, there was a lot of unpredictability regarding which search keywords were generated from the first prompt, especially as the prompt length increased. For best results, stick to shorter prompts. Conclusion In summary, leveraging RAG by combining models allows for more effective finding and sorting of relevant results and generating agent outputs that are far less prone to hallucinations. The smaller the searchable context, the more precise and accurate the responses. Constructing your own custom pipeline of LLM calls offers far more flexibility in tuning responses toward your desired level of accuracy and guard rails. vector and lexical search While it cannot process excessively large amounts of data within the limited context window, it does offer the ability to create effective assistants on limited knowledge bases, as well as run more concurrent agents on the same or lesser hardware than before. This could open up more possibilities for virtual assistants for applications such as tabletop gaming or even cover more complex topics for use by government, legal and accounting firms, scientific research, energy, and more. If you’re ready to get started building, you can . Create your database and start loading your RAG sources today, with no cloud or database Ops experience required. try Astra DB for free By Mario Charnell-Delgado, DataStax Also published . here

Want to Learn Warhammer 40K? Build This Chatbot with Vectors and RAG on Commodity Hardware

Warhammer 40K শিখতে চান? কমোডিটি হার্ডওয়্যারে ভেক্টর এবং RAG দিয়ে এই চ্যাটবট তৈরি করুন

Möchten Sie Warhammer 40K lernen? Erstellen Sie diesen Chatbot mit Vektoren und RAG auf Commodity-Hardware

¿Quieres aprender Warhammer 40K? Cree este chatbot con vectores y RAG en hardware básico

Vous voulez apprendre Warhammer 40K ? Construisez ce chatbot avec des vecteurs et RAG sur du matériel de base

क्या आप वॉरहैमर 40K सीखना चाहते हैं? कमोडिटी हार्डवेयर पर वेक्टर और आरएजी के साथ इस चैटबॉट का निर्माण करें

ウォーハンマー 40K を学びたいですか? Vectors と RAG を使用してコモディティ ハードウェア上でこのチャットボットを構築する

워해머 40K를 배우고 싶으십니까? 상용 하드웨어에서 벡터와 RAG를 사용하여 챗봇 구축

Quer aprender Warhammer 40K? Construa este chatbot com vetores e RAG em hardware de commodities

Хотите изучить Warhammer 40K? Создайте этого чат-бота с помощью Vector и RAG на стандартном оборудовании

Warhammer 40K'yı Öğrenmek İster misiniz? Ticari Donanımda Vektörler ve RAG ile Bu Chatbot'u Oluşturun

Bạn muốn học Warhammer 40K? Xây dựng Chatbot này với Vector và RAG trên phần cứng hàng hóa

想学习战锤40K吗？在商品硬件上使用向量和 RAG 构建此聊天机器人

This paper is available on arxiv under CC 4.0 license. Authors: (1) Berrenur Saylam; (2) Ozlem Durmaz ¨ ˙Incel. Table of Links Highlights Abstract & Introduction Federated Learning: Collaborative and Privacy-Preserving Model Training State of the Art Federated Learning Software Platforms and Testbeds Sensors and Edge-Sensing Devices Federated Learning Applications on Sensing Devices Discussion Conclusions, Acknowledgment, and References 7. Conclusion This paper has provided an overview of the current state-of-the-art FL methodologies applied to sensing devices, focusing on IoT sensors, mobile devices, and wearables. By examining the challenges and opportunities in applying FL to sensing tasks, the paper aims to inspire further research in this domain. The discussion section has raised several important points for consideration. It highlighted the need to address the asynchrony in FL, leverage unlabelled on-device data, explore FL applications in Human Activity Recognition (HAR) and affective computing, and adapt FL methodologies from simulation environments to real-world applications on sensing devices. Efficient wireless communication, resource efficiency, and personalization are crucial when applying FL methodologies to sensing devices. Future research should focus on developing wireless communication techniques that minimize data exchange, designing algorithms with resource constraints in mind, and enabling privacy-preserving personalization. By addressing these challenges, FL methodologies can be effectively applied in real-world sensing applications, considering sensing devices’ unique constraints and requirements. Moreover, the discussion emphasizes the need for studies that bridge the gap between simulation and real-world deployment, ensuring the practicality and effectiveness of FL methodologies in real-world sensing applications. By adapting FL methodologies to real-world settings, researchers can address additional challenges, such as communication constraints, privacy concerns, and device heterogeneity, and validate the performance of FL in diverse sensing domains. In conclusion, this paper aims to provide a comprehensive review for researchers and practitioners interested in applying FL methodologies to edge sensing devices to contribute to developing intelligent and efficient systems in diverse application domains by highlighting the specific challenges and opportunities associated with these devices. Acknowledgment T¨ubitak Bideb 2211-A academic reward is gratefully acknowledged. This research has been supported by the Bo˘gazi¸ci University Research Fund, project number: 19301P. References S. Dey, A. Mukherjee, H. S. Paul, A. Pal, Challenges of using edge devices in iot computation grids, in: 2013 International Conference on Parallel and Distributed Systems, IEEE, 2013, pp. 564–569. L. Atzori, A. Iera, G. Morabito, The internet of things: A survey, Computer networks 54 (2010) 2787–2805. P. Voigt, A. Von dem Bussche, The eu general data protection regulation (gdpr), A Practical Guide, 1st Ed., Cham: Springer International Publishing 10 (2017) 10–5555. M. Capra, R. Peloso, G. Masera, M. Ruo Roch, M. Martina, Edge computing: A survey on the hardware requirements in the internet of things world, Future Internet 11 (2019) 100. Q. Yang, Y. Liu, T. Chen, Y. Tong, Federated machine learning: Concept and applications, ACM Transactions on Intelligent Systems and Technology (TIST) 10 (2019) 1–19. W. Yu, F. Liang, X. He, W. G. Hatcher, C. Lu, J. Lin, X. Yang, A survey on the edge computing for the internet of things, IEEE access 6 (2017) 6900–6919. D. C. Nguyen, M. Ding, P. N. Pathirana, A. Seneviratne, J. Li, H. V. Poor, Federated learning for internet of things: A comprehensive survey, IEEE Communications Surveys & Tutorials 23 (2021) 1622–1658. M. Aledhari, R. Razzak, R. M. Parizi, F. Saeed, Federated learning: A survey on enabling technologies, protocols, and applications, IEEE Access 8 (2020) 140699–140725. J. Xu, B. S. Glicksberg, C. Su, P. Walker, J. Bian, F. Wang, Federated learning for healthcare informatics, Journal of Healthcare Informatics Research 5 (2021) 1–19. Q. Wu, K. He, X. Chen, Personalized federated learning for intelligent iot applications: A cloud-edge based framework, IEEE Open Journal of the Computer Society 1 (2020) 35–44. E. Sannara, F. Portet, P. Lalanda, V. German, A federated learning aggregation algorithm for pervasive computing: Evaluation and comparison, in: 2021 IEEE International Conference on Pervasive Computing and Communications (PerCom), IEEE, 2021, pp. 1–10. F. Wang, A. Preininger, Ai in health: state of the art, challenges, and future directions, Yearbook of medical informatics 28 (2019) 016–026. N. Rieke, J. Hancox, W. Li, F. Milletari, H. R. Roth, S. Albarqouni, S. Bakas, M. N. Galtier, B. A. Landman, K. Maier-Hein, et al., The future of digital health with federated learning, NPJ digital medicine 3 (2020) 119. M. Aledhari, R. Razzak, R. M. Parizi, F. Saeed, Federated learning: A survey on enabling technologies, protocols, and applications, IEEE Access 8 (2020) 140699–140725. Q. Jing, W. Wang, J. Zhang, H. Tian, K. Chen, Quantifying the performance of federated transfer learning, arXiv preprint arXiv:1912.12795 (2019). J. Yao, M. Mouhoub, A. Manashty, Y. Morgan, et al., Analysis of Model Aggregation Techniques in Federated Learning, Ph.D. thesis, Faculty of Graduate Studies and Research, University of Regina, 2021. P. Kairouz, H. B. McMahan, B. Avent, A. Bellet, M. Bennis, A. N. Bhagoji, K. Bonawitz, Z. Charles, G. Cormode, R. Cummings, et al., Advances and open problems in federated learning, Foundations and Trends® in Machine Learning 14 (2021) 1–210. A. Saeed, F. D. Salim, T. Ozcelebi, J. Lukkien, Federated self-supervised learning of multisensor representations for embedded intelligence, IEEE Internet of Things Journal 8 (2020) 1030–1040. T. Li, A. K. Sahu, A. Talwalkar, V. Smith, Federated learning: Challenges, methods, and future directions, IEEE signal processing magazine 37 (2020) 50–60. J. Koneˇcn`y, H. B. McMahan, F. X. Yu, P. Richt´arik, A. T. Suresh, D. Bacon, Federated learning: Strategies for improving communication efficiency, arXiv preprint arXiv:1610.05492 (2016). K. Wang, R. Mathews, C. Kiddon, H. Eichner, F. Beaufays, D. Ramage, Federated evaluation of on-device personalization, arXiv preprint arXiv:1910.10252 (2019). K. Bonawitz, H. Eichner, W. Grieskamp, D. Huba, A. Ingerman, V. Ivanov, C. Kiddon, J. Koneˇcn`y, S. Mazzocchi, B. McMahan, et al., Towards federated learning at scale: System design, Proceedings of machine learning and systems 1 (2019) 374–388. K. Bonawitz, V. Ivanov, B. Kreuter, A. Marcedone, H. B. McMahan, S. Patel, D. Ramage, A. Segal, K. Seth, Practical secure aggregation for privacypreserving machine learning, in: proceedings of the 2017 ACM SIGSAC Conference on Computer and Communications Security, 2017, pp. 1175– 1191. Z. Zhou, S. Yang, L. Pu, S. Yu, Cefl: Online admission control, data scheduling, and accuracy tuning for cost-efficient federated learning across edge nodes, IEEE Internet of Things Journal 7 (2020) 9341–9356. Y. Jiang, R. Cong, C. Shu, A. Yang, Z. Zhao, G. Min, Federated learning based mobile crowd sensing with unreliable user data, in: 2020 IEEE 22nd International Conference on High Performance Computing and Communications; IEEE 18th International Conference on Smart City; IEEE 6th International Conference on Data Science and Systems (HPCC/SmartCity/DSS), IEEE, 2020, pp. 320–327. B. McMahan, E. Moore, D. Ramage, S. Hampson, B. A. y Arcas, Communication-efficient learning of deep networks from decentralized data, in: Artificial intelligence and statistics, PMLR, 2017, pp. 1273–1282. N. Shlezinger, M. Chen, Y. C. Eldar, H. V. Poor, S. Cui, Uveqfed: Universal vector quantization for federated learning, IEEE Transactions on Signal Processing 69 (2020) 500–514. S. P. Karimireddy, S. Kale, M. Mohri, S. Reddi, S. Stich, A. T. Suresh, Scaffold: Stochastic controlled averaging for federated learning, in: International conference on machine learning, PMLR, 2020, pp. 5132–5143. A. Reisizadeh, A. Mokhtari, H. Hassani, A. Jadbabaie, R. Pedarsani, Fedpaq: A communication-efficient federated learning method with periodic averaging and quantization, in: International Conference on Artificial Intelligence and Statistics, PMLR, 2020, pp. 2021–2031. H. Wang, M. Yurochkin, Y. Sun, D. Papailiopoulos, Y. Khazaeni, Federated learning with matched averaging, arXiv preprint arXiv:2002.06440 (2020). M. Zaharia, A. Konwinski, A. D. Joseph, R. H. Katz, I. Stoica, Improving mapreduce performance in heterogeneous environments., in: Osdi, volume 8, 2008, p. 7. C. Xie, S. Koyejo, I. Gupta, Asynchronous federated optimization, arXiv preprint arXiv:1903.03934 (2019). W. Wu, L. He, W. Lin, R. Mao, C. Maple, S. Jarvis, Safa: A semiasynchronous protocol for fast federated learning with low overhead, IEEE Transactions on Computers 70 (2020) 655–668. M. Duan, D. Liu, X. Chen, Y. Tan, J. Ren, L. Qiao, L. Liang, Astraea: Selfbalancing federated learning for improving classification accuracy of mobile deep learning applications, in: 2019 IEEE 37th international conference on computer design (ICCD), IEEE, 2019, pp. 246–254. J. Liu, J. H. Wang, C. Rong, Y. Xu, T. Yu, J. Wang, Fedpa: An adaptively partial model aggregation strategy in federated learning, Computer Networks 199 (2021) 108468. Y. Gao, L. Liu, X. Zheng, C. Zhang, H. Ma, Federated sensing: Edgecloud elastic collaborative learning for intelligent sensing, IEEE Internet of Things Journal 8 (2021) 11100–11111. I. Erg¨un, H. U. Sami, B. G¨uler, Communication-efficient secure aggregation for federated learning, in: GLOBECOM 2022-2022 IEEE Global Communications Conference, IEEE, 2022, pp. 3881–3886. C. He, S. Li, J. So, X. Zeng, M. Zhang, H. Wang, X. Wang, P. Vepakomma, A. Singh, H. Qiu, et al., Fedml: A research library and benchmark for federated machine learning, arXiv preprint arXiv:2007.13518 (2020). J. Xu, S. Wang, L. Wang, A. C.-C. Yao, Fedcm: Federated learning with client-level momentum, arXiv preprint arXiv:2106.10874 (2021). T. Li, A. K. Sahu, M. Zaheer, M. Sanjabi, A. Talwalkar, V. Smith, Federated optimization in heterogeneous networks, Proceedings of Machine learning and systems 2 (2020) 429–450. T. Semwal, A. Mulay, A. M. Agrawal, Fedperf: A practitioners’ guide to performance of federated learning algorithms (2020). I. Kholod, E. Yanaki, D. Fomichev, E. Shalugin, E. Novikova, E. Filippov, M. Nordlund, Open-source federated learning frameworks for iot: A comparative review and analysis, Sensors 21 (2020) 167. TFF, Tensorflow federated: Machine learning on decentralized data, Accessed: 21 September 2023. https://www.tensorflow.org/federated. S. Caldas, S. M. K. Duddu, P. Wu, T. Li, J. Koneˇcn`y, H. B. McMahan, V. Smith, A. Talwalkar, Leaf: A benchmark for federated settings, arXiv preprint arXiv:1812.01097 (2018). O. Community, Open-source python 3 based library developed by the openmined community, Accessed: 21 September 2023. https://blog. openmined.org/tag/pysyft/. D. J. Beutel, T. Topal, A. Mathur, X. Qiu, J. Fernandez-Marques, Y. Gao, L. Sani, K. H. Li, T. Parcollet, P. P. B. de Gusm˜ao, et al., Flower: A friendly federated learning research framework, arXiv preprint arXiv:2007.14390 (2020). Fedai, An industrial grade federated learning framework, Accessed: 21 September 2023. https://fate.fedai.org/. Intel, Open-source python 3 based library for federated learning developed by the intel, Accessed: 21 September 2023. https://github.com/intel/ openfl. M. Popovic, M. Popovic, I. Kastelan, M. Djukic, S. Ghilezan, A simple python testbed for federated learning algorithms, arXiv preprint arXiv:2305.20027 (2023). H. Cho, A. Mathur, F. Kawsar, Flame: Federated learning across multidevice environments, Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 6 (2022) 1–29. R. Jaren, E. Scott, L. Chapin, The internet of things: an overviewunderstanding the issues and challenges of a more connected world, Internet Soc (2015). R. Want, B. N. Schilit, S. Jenson, Enabling the internet of things, Computer 48 (2015) 28–35. H. U. Rehman, M. Asif, M. Ahmad, Future applications and research challenges of iot, in: 2017 International conference on information and communication technologies (ICICT), IEEE, 2017, pp. 68–74. J. McCann, L. Quinn, S. McGrath, E. O’Connell, Towards the distributed edge–an iot review, in: 2018 12th International Conference on Sensing Technology (ICST), IEEE, 2018, pp. 263–268. J. Feng, C. Rong, F. Sun, D. Guo, Y. Li, Pmf: A privacy-preserving human mobility prediction framework via federated learning, Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 4 (2020) 1–21. K. Sozinov, V. Vlassov, S. Girdzijauskas, Human activity recognition using federated learning, in: 2018 IEEE Intl Conf on Parallel & Distributed Processing with Applications, Ubiquitous Computing & Communications, Big Data & Cloud Computing, Social Computing & Networking, Sustainable Computing & Communications (ISPA/IUCC/BDCloud/SocialCom/SustainCom), IEEE, 2018, pp. 1103– 1111. S. Ek, F. Portet, P. Lalanda, G. Vega, Evaluation of federated learning aggregation algorithms: application to human activity recognition, in: Adjunct proceedings of the 2020 ACM international joint conference on pervasive and ubiquitous computing and proceedings of the 2020 ACM international symposium on wearable computers, 2020, pp. 638–643. C. Srinivasan, B. Rajesh, P. Saikalyan, K. Premsagar, E. S. Yadav, A review on the different types of internet of things (iot), Journal of Advanced Research in Dynamical and Control Systems 11 (2019) 154–158. D. C. Nguyen, M. Ding, P. N. Pathirana, A. Seneviratne, J. Li, H. V. Poor, Federated learning for internet of things: A comprehensive survey, IEEE Communications Surveys & Tutorials 23 (2021) 1622–1658. G. Szegedi, P. Kiss, T. Horv´ath, Evolutionary federated learning on eegdata., in: ITAT, 2019, pp. 71–78. T. Wang, Y. Du, Y. Gong, K.-K. R. Choo, Y. Guo, Applications of federated learning in mobile health: Scoping review, Journal of Medical Internet Research 25 (2023) e43006. G. O. Ogbuabor, J. C. Augusto, R. Moseley, A. van Wyk, Context-aware support for cardiac health monitoring using federated machine learning, in: Artificial Intelligence XXXVIII: 41st SGAI International Conference on Artificial Intelligence, AI 2021, Cambridge, UK, December 14–16, 2021, Proceedings 41, Springer, 2021, pp. 267–281. R. W. Picard, Affective computing, MIT press, 2000 Y. S. Can, C. Ersoy, Privacy-preserving federated deep learning for wearable iot-based biomedical monitoring, ACM Transactions on Internet Technology (TOIT) 21 (2021) 1–17. D.-V. Nguyen, K. Zettsu, Spatially-distributed federated learning of convolutional recurrent neural networks for air pollution prediction, in: 2021 IEEE International Conference on Big Data (Big Data), IEEE, 2021, pp. 3601–3608. S. Chen, G. Long, T. Shen, J. Jiang, Prompt federated learning for weather forecasting: Toward foundation models on meteorological data, arXiv preprint arXiv:2301.09152 (2023). J. So, B. G¨uler, A. S. Avestimehr, Turbo-aggregate: Breaking the quadratic aggregation barrier in secure federated learning, IEEE Journal on Selected Areas in Information Theory 2 (2021) 479–489. J. C. Jiang, B. Kantarci, S. Oktug, T. Soyata, Federated learning in smart city sensing: Challenges and opportunities, Sensors 20 (2020) 6230. Y. Liu, J. James, J. Kang, D. Niyato, S. Zhang, Privacy-preserving traffic flow prediction: A federated learning approach, IEEE Internet of Things Journal 7 (2020) 7751–7763. D. C. Nguyen, M. Ding, P. N. Pathirana, A. Seneviratne, J. Li, H. V. Poor, Federated learning for internet of things: A comprehensive survey, IEEE Communications Surveys & Tutorials 23 (2021) 1622–1658. A. M. Elbir, B. Soner, S. C¸ ¨oleri, D. G¨und¨uz, M. Bennis, Federated learning in vehicular networks, in: 2022 IEEE International Mediterranean Conference on Communications and Networking (MeditCom), IEEE, 2022, pp. 72–77. X. Liang, Y. Liu, T. Chen, M. Liu, Q. Yang, Federated transfer reinforcement learning for autonomous driving, in: Federated and Transfer Learning, Springer, 2022, pp. 357–371. Z. Ge, Z. Song, S. X. Ding, B. Huang, Data mining and analytics in the process industry: The role of machine learning, Ieee Access 5 (2017) 20590– 20616. V. Pruckovskaja, A. Weissenfeld, C. Heistracher, A. Graser, J. Kafka, P. Leputsch, D. Schall, J. Kemnitz, Federated learning for predictive maintenance and quality inspection in industrial applications, arXiv preprint arXiv:2304.11101 (2023). X. Han, H. Yu, H. Gu, Visual inspection with federated learning, in: Image Analysis and Recognition: 16th International Conference, ICIAR 2019, Waterloo, ON, Canada, August 27–29, 2019, Proceedings, Part II 16, Springer, 2019, pp. 52–64. F. Liu, X. Wu, S. Ge, W. Fan, Y. Zou, Federated learning for vision-andlanguage grounding problems, in: Proceedings of the AAAI Conference on Artificial Intelligence, volume 34, 2020, pp. 11572–11579. S. Zhai, X. Jin, L. Wei, H. Luo, M. Cao, Dynamic federated learning for gmec with time-varying wireless link, IEEE Access 9 (2021) 10400–10412. N. I. Mowla, N. H. Tran, I. Doh, K. Chae, Federated learning-based cognitive detection of jamming attack in flying ad-hoc network, IEEE Access 8 (2019) 4338–4350. W. Zhou, Y. Li, S. Chen, B. Ding, Real-time data processing architecture for multi-robots based on differential federated learning, in: 2018 IEEE SmartWorld, Ubiquitous Intelligence & Computing, Advanced & Trusted Computing, Scalable Computing & Communications, Cloud & Big Data Computing, Internet of People and Smart City Innovation (SmartWorld/SCALCOM/UIC/ATC/CBDCom/IOP/SCI), IEEE, 2018, pp. 462–471. B. Liu, L. Wang, M. Liu, C.-Z. Xu, Federated imitation learning: A novel framework for cloud robotic systems with heterogeneous sensor data, IEEE Robotics and Automation Letters 5 (2020) 3509–3516. B. Liu, L. Wang, M. Liu, Lifelong federated reinforcement learning: a learning architecture for navigation in cloud robotic systems, IEEE Robotics and Automation Letters 4 (2019) 4555–4562. M. Fu, Y. Shi, Y. Zhou, Federated learning via unmanned aerial vehicle, IEEE Transactions on Wireless Communications (2023). W. Li, W. Yang, G. Jin, J. Chen, J. Li, R. Huang, Z. Chen, Clustering federated learning for bearing fault diagnosis in aerospace applications with a self-attention mechanism, Aerospace 9 (2022) 516. T. Zeng, O. Semiari, M. Mozaffari, M. Chen, W. Saad, M. Bennis, Federated learning in the sky: Joint power allocation and scheduling with uav swarms, in: ICC 2020-2020 IEEE International Conference on Communications (ICC), IEEE, 2020, pp. 1–6. J. Ebenezer, P. G. Krishna, M. Poojitha, A. V. Krishna, Plant leaf disease detection and classification with cnn and federated learning approach, in: Conference of Innovative Product Design and Intelligent Manufacturing System, Springer, 2022, pp. 513–523. A. Abu-Khadrah, A. M. Ali, M. Jarrah, An amendable multi-function control method using federated learning for smart sensors in agricultural production improvements, ACM Transactions on Sensor Networks (2023). This paper is under a CC 4.0 license. available on Arxiv

An Overview of the Current State-of-the-Art Federated Learning Methodologies

The last ten years have been significant here at . We started our journey in December 2013 as a small team of developers on an outsourced project for a marketing company, a partner of Emaar Properties. The project was a request from one of the co-founders, now CTO and Solution Architect, who had just completed the implementation of indoor navigation kiosks for shopping malls. WebLab Technology Remote-first At the beginning of the project, we rented an office room in a business center located in an industrial zone in one of the cities in Ukraine. After 6 months of work, we received additional funding from the client and needed more people to join the team. It so happened that some of our team members were due to leave at the end of that year: one of the key engineers was moving to France, and the other was returning to Dubai. It was quite difficult to hire engineers with the right skill set, mainly because we were losing out to large companies in terms of office quality and space management. But this is what led us to develop our . At the time, in 2014, it was our big competitive advantage because there was a significant pool of engineers who wanted to work remotely. remote-first hiring strategy After this initial growth period, we realized that technical competence alone was no longer enough to make a hard-hitting team. Our remote-first approach necessitated engineers who could work autonomously were able to make decisions on their own and were ready to take responsibility for the result. Dedicated Teams We realized that working in the format of dedicated teams is the best approach for . It means we can establish teams to work exclusively on projects for a particular company while also expanding its internal staff. This model has a number of important other advantages: WebLab Technology 1. A dedicated team focuses only on the needs of a specific client and is always available to respond quickly to solve their business problems. More than just being available, dedicated teams always show up and deliver for their clients, providing reliable, consistent execution. Full focus on client goals. 2. Our team fully adapts to the client’s corporate culture, following established rituals such as daily stand-ups, weekly planning, or progress-tracking meetings. We immerse ourselves in the company’s business objectives and choose the methods of cooperation that are most appropriate for them. Organic integration with the business. 3. — cloud technologies, front-end development, DevOps, etc. We help our clients find exactly the right specialists for their projects. Ready (quick) access to expert knowledge in various fields 4. . Young companies may not have the right culture or approaches, and methodologies to work with engineering teams. We know lots of ready-to-use techniques like proper Scrum, cultural patterns, engagement activities, etc. Ready to use patterns for performance 5. . The client pays only for the necessary functions, not for the maintenance of an additional office, equipment, hiring/recruitment, training, insurance, accounting, legal, taxes, etc. In addition, the team can be flexibly scaled up as needs change. Cost-effectiveness Eastern Europe offers a vast human resource pool that global companies can tap into. Moreover, Eastern European workers have very high and in-demand skill sets — sometimes higher than in Western countries — combined with comparatively lower wages. For example, the average salary of a Senior Software Engineer in is $58k, in — $60k, in — $64k. In the , this figure reaches $145k per year for exactly the same skills and experience. Ukraine Poland Estonia US Therefore, working with from Eastern Europe is an opportunity to get high returns and professional, motivated engineers at optimal prices — a mutually beneficial relationship for all parties. Senior Dedicated Development Teams Values that unite our team Here at WebLab Technology, we have a strict set of values that inform every decision we make. But we did not form them in one day. Our values are a long evolutionary process that is still ongoing. Values let our teammates, clients, and partners know what to expect from us because when people share common values, it creates trust and predictability in relationships. 1. : Always eager to learn and adapt. Open-mindedness Embracing a modest self-assessment. Humility: 3. Upholding consideration and empathy for others. Respect: 4. Striving for excellence autonomously. Self-driven: 5. Addressing issues with sincerity and courage. Honesty: 6. Treating each project as our own, ensuring its success. Ownership: Cooperation with Shaman as one of our turning points Our dedicated development teams usually work in two formats: they extend the client’s existing teams but have a specific area of responsibility, or they are fully responsible for the entire R&D direction of the client company, as was the case with . Shaman is one of our main clients that we are very proud to work with. The company’s technical component is fully implemented by WebLab. It is now an established business with customers among the world’s leading pharmaceutical corporations, such as , , , . Shaman Novartis AstraZeneca Takeda Galderma Our collaboration with started back in 2015 with the idea of an iOS app for internal use to create smart canvas presentations. These presentations needed to be easy to adapt to the audience, sharable, and simple to design. The app has generated a lot of market traction, and it is now used by global enterprises around the world. It was on the basis of this application that we created a B2E SaaS company, which now employs more than 30 of our engineers. Shaman Currently, Shaman is actively promoting its flagship product, , in the pharmaceutical market. In line with Shaman’s mission, the company seeks to empower Life Science Organizations’ content teams with the best no-code platform available, enabling them to create tailored content for healthcare professionals (HCP) in the most efficient way possible. ShamanGo The recent shift of leading pharmaceutical corporations towards digital strategies has increased the demand for unique content. However, many of these companies do not have time to quickly create and manage huge amounts of information. This is where ShamanGo comes in — its users can generate high-quality omnichannel content in minutes by integrating with the leading platform . Thanks to close cooperation with , ShamanGo has become the optimal solution for accelerating content production in the context of the digital transformation of the pharmaceutical industry. Veeva Veeva Systems Over our years of cooperation with Shaman, we have achieved certain successes that have shaped our work together. Some of our proudest achievements include: ● Scaled the team to 30+ engineers; ● Implemented Scrum and Scrum of Scrums to maintain consistency and motivation across our seven dedicated teams; ● Developed a strategy for advanced quality control (a vital concept in the B2E sector due to its demand for stability) and ; 99.9%+ uptime ● Obtained certification; SOC2 Type 2 ● Created a global infrastructure in the US, Europe, and Asia Pacific ● Achieved a high level of infrastructure automation and quality control; ● Achieved top scores on the Framework methodology; AWS Well-Architected ● Supported the Certification process; Veeva Technical Partnership ● Launched a data-driven decision-making system; ● Implemented integration with various third-party platforms, including Salesforce; AI-strategy We believe that and so it is one of our main goals to stay ahead of the competition in this domain. AI will change the world We started early work with AI in 2019, focusing on routing problems, multiple-objective genetic algorithms, and letters using computer vision. Today, we have more than ten successful AI cases in our portfolio. We moved into more active integration of generative AI technologies into business processes in 2022. This has led us to focus on the development of AI agents and RAG solutions based on models such as GPT-4, Llma2, Stable Diffusion, etc. As in other industries, we can see the benefit AI could bring to software engineering and plan to embrace this change. Our past projects include AI solutions for optimizing workflows in pharmaceutical companies. For example, we developed Magic Copy for , as an AI agent based on GPT-4 that generates optimized email templates following simple input instructions from a user. Shaman Another successful case is the use of computer vision in the FoodTech database, which we designed for OllyFood. For this project, we implemented an automated quality control system for dishes and products based on image analysis. The computer vision-based system analyzes images of dishes, identifies all ingredients, assigns quality scores, and detects up to 80% of deviations before the product reaches the customer. We continue to actively develop AI solutions for various industries. Our goal is not only to keep up with technological progress but also to set trends and shape the future of AI. Optimizing Engineer Teams with Data Lake Metrics: A Strategic Approach Building effective teams that succeed is a difficult task in both sports and business. You may have watched the movie Moneyball, which shows how a baseball team manager started using analytical techniques to select players based on statistics rather than subjective impressions of scouts. This allowed him to build a competitive team with a limited budget. In the IT business and software development teams, similar logic applies. We at WebLab Technology build a team and evaluate performance based on data and metrics, as this kind of information is a cornerstone of our industry. One of the key approaches for us is measurement and automation at all levels: the company as a whole, teams, and staff roles. For example, we have an that tracks the number of completed story points, planned scope of work, standard Scrum reports, the share of unplanned tasks and bugs, business value points, pull request sizes, code integration frequency, QA returned count, etc. automated management system In addition to hard metrics, we also quantitatively analyze : motivation, team atmosphere, and the realization of one’s capabilities. However, career advancement in the company mostly depends on the results of hard metrics. soft metrics Remote working gives us full digitalization and the ability to continuously track all processes in the company — from the work of engineers to customer satisfaction. We make our personnel and strategic decisions based on all this collected data and analytics we have at our disposal. We believe that the best results are achieved by combining technological and human factors. Hiring slowly, hiring right: Why do we prioritize a thorough candidate search? When it comes to hiring new employees at , we avoid rushing things, even if it may delay the start of their work. In our opinion, accelerated recruitment creates significant risks in the long run, which almost always outweigh the potential short-term benefits. That’s why our hiring cycle is almost than the market average, and we pay a lot of attention to assessing the technical skills of potential senior developers. WebLab Technology three times longer The standard search for potential candidates consists of 6 stages: WebLab Technology . We thoroughly review resumes and invite those who initially meet our requirements and have relevant skills. Deep profile screening . Sometimes it’s called a homework assignment. Candidates undergo a specialized Selection Task that evaluates abilities for the specific role. Selection challenge or project The focus is on understanding how the candidate handled specific challenges or tasks in the past, aiming to predict their future performance. Behavioral interview. . Candidates showcase hands-on skills through coding challenges and sample projects. Coding interviews The candidate is typically asked to solve real-world coding problems, demonstrate their coding abilities, and showcase their understanding of relevant technologies within the industry. Domain/coding interview with leading engineer. Our CTO assesses the candidate’s understanding of system design principles, their ability to make informed decisions, and their proficiency in creating scalable and efficient solutions. System design interview with CTO. For ten years, we have honed our ability to recognize the optimal combination of technical skills and the right human qualities, which allows us to expand our dedicated teams with truly motivated and professional engineers. We are also open to hiring new specialists, especially AI Prompting Engineers. If you are an area specialist and think you are a match in terms of values and philosophy, we invite you to check out the vacancies on our website: https://www.weblab.technology/jobs Technology and logic We mainly work with B2B SaaS solutions that require solving specific business problems at scale and primarily focus on logic, functionality, and metrics rather than aesthetic visualization. We have worked with startups at different stages — in their earliest days and when they were in an active growth phase, building up their capacity and staff. Due to our flexibility, we adapt to the needs of our clients at any stage, regardless of the size of the startup and business model. In addition to startups, has implemented complex solutions that require optimization and scaling for global enterprise companies. WebLab Technology Today, our main technical focus is on AWS, logic-intensive micro frontends, microservices, and AI. We also prepare our clients to pass standard certifications and audits and maintain a high level of quality in the areas of security and compliance. For example, we helped our client obtain the . ShamanGo SOC 2 Type II certification Ten years To celebrate the 10th anniversary of our company, we decided not to organize pompous offline events but instead to transfer our celebration budget of more than 10 thousand USD to the Foundation and support those who are now defending Ukraine and the rest of the civilized world. Kolo “The Kolo Foundation is doing a great job. We’ve been cooperating with the foundation for a while now, and I think it’s the best use of our team’s celebration budget,” says WebLab Technology CTO and Solution Architect . Oleksandr Knyga In 10 years, has grown ten-fold. Today, our team consists of professional engineers working on products of varying complexity, and the company has in its portfolio such clients as , , , , . WebLab Technology ShamanGo AstraZeneca Novartis Bryj (ex. FollowAnalytics) Emaar Properties We constantly develop specialists within dedicated teams and share expertise between them for a synergistic effect. Unlike classical outsourcing companies, we are primarily a team of strong engineers, where even the CEO is ready to commit to the repository. Our development focuses on an increase in the average level of seniority of engineers rather than quantitative growth.

Celebrating 10 years of WebLab Technology: Our Story of Growing Through Dedicated Teams

ওয়েবল্যাব প্রযুক্তির 10 বছর উদযাপন: আমাদের ডেডিকেটেড টিমের মাধ্যমে বেড়ে ওঠার গল্প

Wir feiern 10 Jahre WebLab-Technologie: Unsere Geschichte des Wachstums durch engagierte Teams

Celebrando 10 años de tecnología WebLab: nuestra historia de crecimiento a través de equipos dedicados

Célébration des 10 ans de la technologie WebLab : notre histoire de croissance grâce à des équipes dédiées

वेबलैब टेक्नोलॉजी के 10 साल पूरे होने का जश्न: समर्पित टीमों के माध्यम से आगे बढ़ने की हमारी कहानी

WebLab テクノロジーの 10 周年を祝う: 専任チームによる成長のストーリー

WebLab 기술 10주년 기념: 전담 팀을 통한 성장 이야기

Comemorando 10 anos de tecnologia WebLab: nossa história de crescimento por meio de equipes dedicadas

Празднование 10-летия технологии WebLab: наша история роста благодаря преданным командам

WebLab Teknolojisinin 10. yılını kutluyoruz: Özel Ekiplerle Büyüme Hikayemiz

Kỷ niệm 10 năm Công nghệ WebLab: Câu chuyện phát triển của chúng tôi thông qua các nhóm tận tâm

庆祝 WebLab 技术 10 周年：我们通过专业团队成长的故事

Stories, News & Mentions about ASTRA

Astra’s Stories on HackerNoon

Astra's News & Mentions