paint-brush
Neo4j Is Building an Ecosystem of Graph-powered Features for Generative AIby@linked_do
301 reads
301 reads

Neo4j Is Building an Ecosystem of Graph-powered Features for Generative AI

by George AnadiotisMarch 28th, 2024
Read on Terminal Reader
Read this story w/o Javascript
tldt arrow

Too Long; Didn't Read

From better together to full native integration, Neo4j is creating an ecosystem around all major cloud platforms to enable provide graph-powered features for Generative AI and beyond. Here’s how this aligns with cloud platform AI strategies and what’s next. Hint: Databricks and Snowflake.
featured image - Neo4j Is Building an Ecosystem of Graph-powered Features for Generative AI
George Anadiotis HackerNoon profile picture

From better together to full native integration, Neo4j is creating an ecosystem around all major cloud platforms to provide graph-powered features for Generative AI and beyond. Here’s how this aligns with cloud platform AI strategies and what’s next. Hint: Databricks and Snowflake.


For many organizations today, data management comes down to handing over their data to one of the “Big 5” data vendors: AWS, Microsoft Azure, and Google Cloud Platform, plus Databricks and Snowflake. Whether that’s a good idea or not, or whether there’s a chance of a “sixth data platform” emerging, is a different story.


The reality is that the “Big 5” is where a significant portion of the world’s data lives and, consequently, where many analytics and AI-powered applications live as well. If you are a data and application development platform and you want to get in front of as many users as possible, making your platform available via native integration with the “Big 5” and providing value-added capabilities for AI-powered applications seems like a good idea.


That’s precisely how graph database and analytics provider Neo4j’s strategy and roadmap evolved in 2023, as shared by Neo4j Chief Product Officer Sudhir Hasbe. Neo4j started with Google Cloud and then moved to AWS. Now, Neo4j is announcing a strategic collaboration with Microsoft to provide graph-powered features for Generative AI and beyond.


We caught up with Hasbe to discuss what the partnership with Microsoft entails, how it works, and what it enables, as well as how Neo4j’s strategy is evolving and what the landscape looks like.

Neo4j Integrates with Microsoft to Evolve Data Analytics and Enhance GenAI

Neo4j announced the general availability of Neo4j’s fully managed graph database offering AuraDB on the Azure Marketplace, as well as integration with Microsoft Fabric and Microsoft Azure OpenAI. The idea is to enable users to access even more hidden patterns and relationships contained within the new data sets they can access using the Neo4j graph database.


The integration focuses on the following features:

  1. Transforming unstructured data into knowledge graphs
  2. Enhancing contextual understanding and explainability with GraphRAG
  3. Providing a long-term memory for LLMs with vector embedding integration
  4. Enabling graph-powered insights as part of Microsoft Fabric unified data platform
  5. Delivering graph analytics as a native Fabric workload


The integration with Azure OpenAI Service is generally available now, while the Microsoft Fabric integration (i.e. the last 2 items in this list) will be generally available later this year. Hasbe shared a demo, and referred to this as a fully integrated experience inside the Fabric platform that Neo4j is building together with Microsoft.

This, Hasbe explained, is strategically important for Neo4j. Microsoft has more than 350,000 customers using Power BI, making it one of the largest analytics platforms. What’s even more important, he added, is that this is a “democratized” platform that lowers the bar to accessing analytics beyond the technically savvy. By becoming a part of it, Neo4j becomes accessible to a wider audience.

Graph-powered features for Generative AI across cloud platforms

This seems like a win-win for both Neo4j and Microsoft. Previously, Neo4j had also implemented similar integrations with Google Cloud Platform and AWS. In fact, comparing the announcements side by side, similarities abound. Even the technical architecture diagrams look similar, and that’s not that strange.


Hasbe noted that the capabilities that Neo4j wants to have are fully integrated at the base level in all three cloud providers, and they’re equal. The terminology may slightly differ, as Microsoft made GraphRAG quite prominent.


However, transforming unstructured data into knowledge graphs, providing a long-term memory for LLMs with vector embedding integration and enhancing contextual understanding and explainability with GraphRAG are all things that Neo4j has been working on, and promoting, for a while now. It makes sense to have them shared across the board.

Neo4j’s Knowledge Graph and Generative AI reference architecture


Neo4j’s Knowledge Graph and Generative AI reference architecture. This looks remarkably similar across cloud platforms.

But there are also differences, as every cloud provider has some unique capabilities. Hasbe noted that the extensibility points that Microsoft Fabric provides are much stronger than those of any other platform. That’s what enables Neo4j to target different user personalities.


Historically, Neo4j has targeted developers and some high end analysts. But by making Neo4j available as a simple, targeted experience next to Power BI, graph analytics can be democratized. This is why Hasbe is most excited about this integration, even though as he acknowledged there are a number of challenges around that, both from a technical as well as from a go to market perspective.

Graph as part of cloud platforms strategy

Besides Fabric’s extensibility points, there may also be also strategic reasons why the integration with Microsoft is deeper. AWS does have its own graph database platform, Amazon Neptune. Neo4j and Amazon Neptune have been on similar trajectories, so there’s a bit of co-opetition going on there. This is not at all unusual for any vendor working with the hyperscalers, however things are different for Google and Microsoft.


Google does not have a product in the graph database market. Microsoft is sporting its own CosmosDB, which provides graph analytics capabilities. However, despite its merits, CosmosDB is more of a multi-model database with an add-on graph API than a bona fide graph database. While CosmosDB may also support vector embeddings, it’s hard to see how it could fully support GraphRAG.


Hasbe also concurred that Microsoft is investing strategically in AI on all levels, and GraphRAG is a part of this on which Neo4j is very much aligned. As far as Neo4j’s own AI strategy, Hasbe described it along three key lines.


The first one is being part of the ecosystem of AI platforms. Making sure that Neo4j is inherently integrated with them is the most important thing, as he put it. The second one is providing AI experiences for Neo4j customers, such as copilots or natural language query interfaces. Last but not least, Neo4j also uses AI capabilities internally in areas such as log analysis and marketing.

Neo4’s roadmap, Databricks and Snowflake

As far as Neo4j’s overall strategy goes, Hasbe affirmed the progress made on the roadmap shared in 2023. Scalability has been upgraded, and customers are now looking at working with “tens to hundreds of terabytes of large graphs.” As the capability to work with unstructured data is added, graph sizes are going to become bigger and bigger.


There are a number of engineering and usability improvements that Neo4j is working on, with a focus on the cloud platform which is seeing the most traction. Making it as easy as possible to get started with graph coming from relational or unstructured data, also leveraging AI, is a big item in the list.


Equally important, Hasbe said there is a lot of focus this year on ecosystems and how to integrate and got to market with them. Databricks and Snowflake have been in Neo4j’s list for a while, and there will more announcements in the coming months.


Also published here.