Real Madrid&#x27;s squad list for Friday&#x27;s La Liga clash against Real Sociedad has been changed at the last minute in a surprise development.

Real Madrid Squad For Real Sociedad Has Surprise Last Minute Changes

From the wealth of Double Gameweek options last week to some unconvincing &#x27;doublers&#x27; this time: the Scout Squad returns for Gameweek 35.

The Scout Squad’s best players for FPL Gameweek 35

The English midfielder has been ruled out of Real Madrid&#x27;s LaLiga game against La Real. Here&#x27;s the latest.

Is Jude Bellingham injured? Why is he not in the squad against Real Sociedad? - AS USA

VentureBeat is the leader in covering transformative tech. We help business leaders make smarter decisions with our industry-leading AI and gaming coverage.

Supercell to release Squad Busters, 1st new game in over 5 years

England have big decisions to make as they prepare to name their squad for the T20 World Cup, says Stephan Shemilt.

Men's T20 World Cup: England prepare to name squad - BBC Sport

As the deadline for the selection of India&#x27;s T20 World Cup 2024 squad nears, several experts have shared their views on the same.

Mayank Yadav, Riyan Parag In, No Hardik Pandya, Rishabh Pant In T20 World Cup Squad By 6-Time IPL Winner | Cricket News

A police incident has been reported near Oxenholme station

RECAP: Bomb squad called after 'suspicious' item found | The Westmorland Gazette

NBC News projects that Lee, a progressive member of the “Squad,” defeated fellow Democrat Bhavini Patel in Pennsylvania’s 12th District primary.

Rep. Summer Lee fends off a primary challenge in a race that tested Democratic sentiment over Israel

Cricket News: Former Indian cricketers Irfan Pathan and Varun Aaron have suggested the inclusion of MS Dhoni in India&#x27;s T20 World Cup squad, despite his retirement

MS Dhoni: 'No one will mind': Former India cricketers float idea of MS Dhoni's 'wildcard' entry in T20 World Cup squad | Cricket News - Times of India

The former president’s win — which was so decisive that the AP projected him the victor solely based on a voter survey — came in the home state of his only major Republican rival, Nikki Haley · By Nahal Toosi, Ari Hawkins, Koen Verhelst, Gabriel Gavin and Kyle Duggan

Summer Lee of the 'Squad' beats back primary challenge

Sqwad

sqwad

Who would have thought that one of the most pervasive scams on the internet would involve the infamous Geek Squad? The Geek Squad scam is specifically designed to trick you into giving up personal information so the scammer can empty as much of your bank account as possible. This scam can trick anyone unless you know what signs to look for. This post will show you all the dirty tricks used in this scam, so you don’t have to be another victim. What is Geek Squad? If you happen to be a “person in your advance years” wondering why you got a random email from something called “The Geek Squad,” here’s a quick rundown of what this company is. The Geek Squad is a subsidiary of that focuses on support and hardware repair. Basically, it’s a group of helpful nerds that can save you a whole lot of time by fixing your devices. Best Buy tech Does Geek Squad Have A Subscription Service? Yes, Geek Squad does, in fact, have an annual subscription service. For $199.99 per year, you can purchase , a 24/7 unlimited technology support service. This service can save you a lot of money if you constantly have problems with your computer or devices. Geek Squad’s Total Tech Support Unfortunately, this money-saving service is attracting a lot of scammers. These scams are working on victims of all ages, and people are getting tired of it. The scam involves tricking people into thinking that Geek Squad took money out of their accounts. How the Geek Squad Scam Works The Geek Squad email scam centers around renewing their unlimited tech service. There are different versions of this scam, but they’ll usually look something like this: The victim receives an email from the “Geek Squad” claiming that their subscription will be renewed automatically. The price for the renewal is around $400, which is twice the amount it should be. A link or a phone number is provided at the end of the email, with an urgent message to click or call if there are any problems with the price. If the victim clicks the link, malicious malware is downloaded to their computer, with the goal of stealing their identity/sensitive information. If the victim calls the number, the victim is instructed to download software to fix the problem, which adds malware to the system. If you’ve been fooled by one of these emails, don’t feel bad about it. Scammers do an excellent job at making these emails look real by adding official logos, promotional banners, renewal dates, and other order details. Here are two quick examples: How to Quickly Identify a Geek Squad Email Scam No fake email from “Geek Squad” looks the same, but there are similarities between all of them. Once you know what to look for, you’ll never be fooled again. Here are three Geek Squad email examples to best explain how to avoid this scam. Geek Squad Renewal Scam Example #1 This example looks very professional at first glance, and it’s easy to see why some people might get tricked. But with a closer look, there are some immediate red flags here. If this email was really from Geek Squad, it would address the reader directly with their name. This opening is generic, and the “ ” in “ ” needs to be capitalized. c customer There needs to be a space between “ ” and “ .” placed membership The service that Geek Squad provides is not a “ .” “ ” is used here as a buzz term to add legitimacy to the email. malware account Malware “ ” is abbreviated for some reason. This is a big slip-up from the scammer, as a real company would be more professional. Account Here’s a big one that will be present for most of these fake emails. Scammers like to create a sense of urgency so the victim doesn’t have time to think about what they’re doing. Phrases like “ ” create that sense of urgency. Geek Squad, like most companies, give their customers up to 7 days to cancel their subscriptions. within the next 24 hours Also, in this example, “ ” is missing in the phrase, and “ ” would normally be spelled out as “ ” if this was real. the hrs hours Geek Squad Renewal Scam Example #2 Here’s an example that should be an obvious fake for most people. But let me break this down to show how often the same red flags show up in these fake emails. First, let me get the elephant in the room out of the way. The amount of typos in this email is mind-boggling. Not only that, but the text is constantly switching to random capital letters, bold text, and underlines. I won’t cover every typo in this email, as we’ll be here all day. Instead, here are the big red flags you’ll come across often: The email should say it’s from Geek Squad itself, not a random person. \ Not only does this email not address the recipient of the email, but “ ” is just plain weird. This email is not from a native English speaker. Hey True Self This one is a little different from a basic typo. Grammar issues are usually a red flag, as professional companies hire professional copywriters to produce these emails. In this case, “ ” and “ should be capitalized. A mistake even the worst copywriter would catch. thanks since” Here’s one that’s easy to miss. The format of this date is correct in some places in the world, but in America and other countries, the month should come before the day. If there’s a date on the email, make sure it’s in the format of your country. Geek Squad Renewal Scam Example #3 Here’s the toughest example in the bunch. Not only does this email look professional, but the text itself is almost impeccable compared to other fake emails. Still, the red flags are blaring. The chances of a real company charging you without adequate warning are very low. Also, the grammar ￼in this first line is a little off. Plus, the word “ ” is used twice in the same sentence, which is not the best writing. today Here’s that sense of urgency again. If this email were real, you would get more than 24 hours. \ This is probably the most prominent typo in the email. This sentence needs a few more words to make it more cohesive. The fake phone number is in red, which is supposed to grab your attention, but a real company is unlikely to try something so amateur.￼ The service is called . Total Tech Support At the end of the email, there’s a call to action for the reader to call the number again. This is a desperate attempt to add more urgency. It’s sad, really. Tips For Dealing With The Geek Squad Email Scam Receiving one of these fake Geek Squad renewal scam emails is not the end of the world. Here’s what to do if you receive one: Ask yourself if you’ve visited or purchased anything at Best Buy recently. If not, then it doesn’t make sense for Geek Squad to contact you. Check the currency. If the currency doesn’t match the part of the world you live in, then you probably don’t have anything to worry about. Check your credit card or bank account to see if any charges have been made. This will give you peace of mind once you see that nothing has been taken out. Don't send personal information through email or text (bank account, credit card, date of birth). If the email was really from Geek Squad, they would already have that information. Don’t click on any links. Don’t call any phone numbers in the email. Advice From Best Buy After Getting Scammed This scam is so big that had to address the problem themselves. Best Buy Best Buy’s official advice to victims of this scam is to report the email to local law enforcement, as well as the and the . FTC internet crime complaint center Best Buy also suggests that victims change any passwords that scammers may have gotten hold of and check their bank or credit card company for unauthorized purchases. Conclusion In the near future, scammers will use AI to write fake emails, eliminating easy-to-catch typos from the equation. But don’t worry. You can still avoid the Geek Squad Scam by looking for the red flags that will always be there, such as an impersonal greeting, strange date formats, or a warning designed to promote a sense of urgency. Just remember not to call the phone number, click on any links or give out your personal information.

How the Geek Squad Scam Works  (Don't Fall for This)

গিক স্কোয়াড স্ক্যাম কীভাবে কাজ করে (এর জন্য পড়বেন না)

Wie der Geek Squad-Betrug funktioniert (fallen Sie nicht darauf herein)

Cómo funciona la estafa de Geek Squad (no se deje engañar por esto)

Comment fonctionne l'escroquerie Geek Squad (ne tombez pas dans le piège)

गीक स्क्वाड घोटाला कैसे काम करता है (इसके चक्कर में न पड़ें)

Geek Squad 詐欺の仕組み (騙されないでください)

Geek Squad 사기의 작동 방식(이것에 속지 마세요)

Como funciona o golpe do Geek Squad (não caia nessa)

Geek Squad Dolandırıcılığı Nasıl Çalışır (Buna Kanmayın)

Cách thức hoạt động của Geek Squad Scam (Đừng sa ngã vì điều này)

极客小队骗局是如何运作的（不要上当）

The Mad King by Edgar Rice Burroughs, is part of the HackerNoon Books Series. You can jump to any chapter in this book here . BEFORE THE FIRING SQUAD BEFORE THE FIRING SQUAD They marched Barney before the staff where he urged his American nationality, pointing to his credentials and passes in support of his contention. The general before whom he had been brought shrugged his shoulders. “They are all Americans as soon as they are caught,” he said; “but why did you not claim to be Prince Peter of Blentz? You have his passes as well. How can you expect us to believe your story when you have in your possession passes for different men? “We have every respect for our friends the Americans. I would even stretch a point rather than chance harming an American; but you will admit that the evidence is all against you. You were found in the very building where Drontoff was known to stay while in Burgova. The young woman whose mother keeps the place directed our officer to your room, and you tried to escape, which I do not think that an innocent American would have done. “However, as I have said, I will go to almost any length rather than chance a mistake in the case of one who from his appearance might pass more readily for an American than a Serbian. I have sent for Prince Peter of Blentz. If you can satisfactorily explain to him how you chance to be in possession of military passes bearing his name I shall be very glad to give you the benefit of every other doubt.” Peter of Blentz. Send for Peter of Blentz! Barney wondered just what kind of a sensation it was to stand facing a firing squad. He hoped that his knees wouldn’t tremble—they felt a trifle weak even now. There was a chance that the man might not recall his face, but a very slight chance. It had been his remarkable likeness to Leopold of Lutha that had resulted in the snatching of a crown from Prince Peter’s head. Likely indeed that he would ever forget his, Barney’s, face, though he had seen it but once without the red beard that had so added to Barney’s likeness to the king. But Maenck would be along, of course, and Maenck would have no doubts—he had seen Barney too recently in Beatrice to fail to recognize him now. Several men were entering the room where Barney stood before the general and his staff. A glance revealed to the prisoner that Peter of Blentz had come, and with him Von Coblich and Maenck. At the same instant Peter’s eyes met Barney’s, and the former, white and wide-eyed came almost to a dead halt, grasping hurriedly at the arm of Maenck who walked beside him. “My God!” was all that Barney heard him say, but he spoke a name that the American did not hear. Maenck also looked his surprise, but his expression was suddenly changed to one of malevolent cunning and gratification. He turned toward Prince Peter with a few low-whispered words. A look of relief crossed the face of the Blentz prince. “You appear to know the gentleman,” said the general who had been conducting Barney’s examination. “He has been arrested as a Serbian spy, and military passes in your name were found upon his person together with the papers of an American newspaper correspondent, which he claims to be. He is charged with being Stefan Drontoff, whom we long have been anxious to apprehend. Do you chance to know anything about him, Prince Peter?” “Yes,” replied Peter of Blentz, “I know him well by sight. He entered my room last night and stole the military passes from my coat—we all saw him and pursued him, but he got away in the dark. There can be no doubt but that he is the Serbian spy.” “He insists that he is Bernard Custer, an American,” urged the general, who, it seemed to Barney, was anxious to make no mistake, and to give the prisoner every reasonable chance—a state of mind that rather surprised him in a European military chieftain, all of whom appeared to share the popular obsession regarding the prevalence of spies. “Pardon me, general,” interrupted Maenck. “I am well acquainted with Mr. Custer, who spent some time in Lutha a couple of years ago. This man is not he.” “That is sufficient, gentlemen, I thank you,” said the general. He did not again look at the prisoner, but turned to a lieutenant who stood near-by. “You may remove the prisoner,” he directed. “He will be destroyed with the others—here is the order,” and he handed the subaltern a printed form upon which many names were filled in and at the bottom of which the general had just signed his own. It had evidently been waiting the outcome of the examination of Stefan Drontoff. Surrounded by soldiers, Barney Custer walked from the presence of the military court. It was to him as though he moved in a strange world of dreams. He saw the look of satisfaction upon the face of Peter of Blentz as he passed him, and the open sneer of Maenck. As yet he did not fully realize what it all meant—that he was marching to his death! For the last time he was looking upon the faces of his fellow men; for the last time he had seen the sun rise, never again to see it set. He was to be “destroyed.” He had heard that expression used many times in connection with useless horses, or vicious dogs. Mechanically he drew a cigarette from his pocket and lighted it. There was no bravado in the act. On the contrary it was done almost unconsciously. The soldiers marched him through the streets of Burgova. The men were entirely impassive—even so early in the war they had become accustomed to this grim duty. The young officer who commanded them was more nervous than the prisoner—it was his first detail with a firing squad. He looked wonderingly at Barney, expecting momentarily to see the man collapse, or at least show some sign of terror at his close impending fate; but the American walked silently toward his death, puffing leisurely at his cigarette. At last, after what seemed a long time, his guard turned in at a large gateway in a brick wall surrounding a factory. As they entered Barney saw twenty or thirty men in civilian dress, guarded by a dozen infantrymen. They were standing before the wall of a low brick building. Barney noticed that there were no windows in the wall. It suddenly occurred to him that there was something peculiarly grim and sinister in the appearance of the dead, blank surface of weather-stained brick. For the first time since he had faced the military court he awakened to a full realization of what it all meant to him—he was going to be lined up against that ominous brick wall with these other men—they were going to shoot them. A momentary madness seized him. He looked about upon the other prisoners and guards. A sudden break for liberty might give him temporary respite. He could seize a rifle from the nearest soldier, and at least have the satisfaction of selling his life dearly. As he looked he saw more soldiers entering the factory yard. A sudden apathy overwhelmed him. What was the use? He could not escape. Why should he wish to kill these soldiers? It was not they who were responsible for his plight—they were but obeying orders. The close presence of death made life seem very desirable. These men, too, desired life. Why should he take it from them uselessly? At best he might kill one or two, but in the end he would be killed as surely as though he took his place before the brick wall with the others. He noticed now that these others evinced no inclination to contest their fates. Why should he, then? Doubtless many of them were as innocent as he, and all loved life as well. He saw that several were weeping silently. Others stood with bowed heads gazing at the hard-packed earth of the factory yard. Ah, what visions were their eyes beholding for the last time! What memories of happy firesides! What dear, loved faces were limned upon that sordid clay! His reveries were interrupted by the hoarse voice of a sergeant, breaking rudely in upon the silence and the dumb terror. The fellow was herding the prisoners into position. When he was done Barney found himself in the front rank of the little, hopeless band. Opposite them, at a few paces, stood the firing squad, their gun butts resting upon the ground. The young lieutenant stood at one side. He issued some instructions in a low tone, then he raised his voice. “Ready!” he commanded. Fascinated by the horror of it, Barney watched the rifles raised smartly to the soldiers’ hips—the movement was as precise as though the men were upon parade. Every bolt clicked in unison with its fellows. “Aim!” the pieces leaped to the hollows of the men’s shoulders. The leveled barrels were upon a line with the breasts of the condemned. A man at Barney’s right moaned. Another sobbed. “Fire!” There was the hideous roar of the volley. Barney Custer crumpled forward to the ground, and three bodies fell upon his. A moment later there was a second volley—all had not fallen at the first. Then the soldiers came among the bodies, searching for signs of life; but evidently the two volleys had done their work. The sergeant formed his men in line. The lieutenant marched them away. Only silence remained on guard above the pitiful dead in the factory yard. The day wore on and still the stiffening corpses lay where they had fallen. Twilight came and then darkness. A head appeared above the top of the wall that had enclosed the grounds. Eyes peered through the night and keen ears listened for any sign of life within. At last, evidently satisfied that the place was deserted, a man crawled over the summit of the wall and dropped to the ground within. Here again he paused, peering and listening. What strange business had he here among the dead that demanded such caution in its pursuit? Presently he advanced toward the pile of corpses. Quickly he tore open coats and searched pockets. He ran his fingers along the fingers of the dead. Two rings had rewarded his search and he was busy with a third that encircled the finger of a body that lay beneath three others. It would not come off. He pulled and tugged, and then he drew a knife from his pocket. But he did not sever the digit. Instead he shrank back with a muffled scream of terror. The corpse that he would have mutilated had staggered suddenly to its feet, flinging the dead bodies to one side as it rose. “You fiend!” broke from the lips of the dead man, and the ghoul turned and fled, gibbering in his fright. The tramp of soldiers in the street beyond ceased suddenly at the sound from within the factory yard. It was a detail of the guard marching to the relief of sentries. A moment later the gates swung open and a score of soldiers entered. They saw a figure dodging toward the wall a dozen paces from them, but they did not see the other that ran swiftly around the corner of the factory. This other was Barney Custer of Beatrice. When the command to fire had been given to the squad of riflemen, a single bullet had creased the top of his head, stunning him. All day he had lain there unconscious. It had been the tugging of the ghoul at his ring that had roused him to life at last. Behind him, as he scurried around the end of the factory building, he heard the scattering fire of half a dozen rifles, followed by a scream—the fleeing hyena had been hit. Barney crouched in the shadow of a pile of junk. He heard the voices of soldiers as they gathered about the wounded man, questioning him, and a moment later the imperious tones of an officer issuing instructions to his men to search the yard. That he must be discovered seemed a certainty to the American. He crouched further back in the shadows close to the wall, stepping with the utmost caution. Presently to his chagrin his foot touched the metal cover of a manhole; there was a resultant rattling that smote upon Barney’s ears and nerves with all the hideous clatter of a boiler shop. He halted, petrified, for an instant. He was no coward, but after being so near death, life had never looked more inviting, and he knew that to be discovered meant certain extinction this time. The soldiers were circling the building. Already he could hear them nearing his position. In another moment they would round the corner of the building and be upon him. For an instant he contemplated a bold rush for the fence. In fact, he had gathered himself for the leaping start and the quick sprint across the open under the noses of the soldiers who still remained beside the dying ghoul, when his mind suddenly reverted to the manhole beneath his feet. Here lay a hiding place, at least until the soldiers had departed. Barney stooped and raised the heavy lid, sliding it to one side. How deep was the black chasm beneath he could not even guess. Doubtless it led into a coal bunker, or it might open over a pit of great depth. There was no way to discover other than to plumb the abyss with his body. Above was death—below, a chance of safety. The soldiers were quite close when Barney lowered himself through the manhole. Clinging with his fingers to the upper edge his feet still swung in space. How far beneath was the bottom? He heard the scraping of the heavy shoes of the searchers close above him, and then he closed his eyes, released the grasp of his fingers, and dropped. About HackerNoon Book Series: We bring you the most important technical, scientific, and insightful public domain books. This book is part of the public domain. Edgar Rice Burroughs (1995). The Mad King. Urbana, Illinois: Project Gutenberg. Retrieved October 2022 https://www.gutenberg.org/cache/epub/364/pg364-images.html This eBook is for the use of anyone anywhere at no cost and with almost no restrictions whatsoever. You may copy it, give it away or re-use it under the terms of the Project Gutenberg License included with this eBook or online at www.gutenberg.org , located at https://www.gutenberg.org/policy/license.html.

BEFORE THE FIRING SQUAD

The Ray Shot Down and Squadrons of Planes Frizzled Like Moths in the Air

El rayo derribado y los escuadrones de aviones se encresparon como polillas en el aire

Le rayon abattu et les escadrons d'avions frisés comme des papillons de nuit dans les airs

रे शॉट डाउन और विमानों के स्क्वाड्रन हवा में पतंगों की तरह जम गए

光線が撃ち落とされ、飛行機の飛行隊が空中の蛾のように縮れた

O raio abatido e esquadrões de aviões frisados como mariposas no ar

Tia bắn hạ và Phi đội Máy bay Lăn tăn như Bướm đêm trên không

射线被击落，飞机中队像空中飞蛾一样卷曲

Open sourced by Google Research team, achieved wide popularity amongst NLP enthusiasts for all the right reasons! It is one of the best Natural Language Processing pre-trained models with superior NLP capabilities. It can be used for language classification, question & answering, next word prediction, tokenization, etc. pre-trained models of BERT Our case study Question Answering System in Python using BERT NLP and BERT based Question and Answering system demo , developed in Python + Flask, got hugely popular garnering hundreds of visitors per day. We got a lot of appreciative and lauding emails praising our QnA demo. Along with that, we also got number of people asking about how we created this QnA demo. And till the day, we keep getting requests on how to develop such a QnA system using BERT pre-trained model open-sourced by Google. [1] [2] To start with, the readme file on the official GitHub repository of BERT provides a good amount of information about how to fine-tune the model on SQuAD 2.0 but we could see that developers are still facing issues. So, we decided to publish a step-by-step tutorial to fine-tune the BERT pre-trained model and generate inference of answers from the given paragraph and questions on using TPU. Colab In this tutorial, we are not going to cover how to create web-based interface using Python + Flask. We’ll just cover the fine-tuning and inference on Colab using TPU. You can create your own interface using Flask or Django. Overview In this tutorial we will see how to perform a fine-tuning task on SQuAD using Google Colab, for that we will use BERT GitHub Repository, BERT Repository includes: TensorFlow code for the BERT model architecture. Pre-trained models for both the lowercase and cased version of BERT-Base and BERT-Large. You can also refer or copy to follow the steps. our colab file Steps to perform BERT Fine-tuning on Google Colab 1) Change Runtime to TPU On the main menu, click on and select . Set “ “ as the hardware accelerator. Below screeenshot will help you understand how you can change the runtime to TPU. Runtime Change runtime type TPU After Clicking on “Change runtime type”, Select from the dropdown option as given in the below figure. TPU 2) Clone the BERT github repository , or idirectional mbedding epresentations from ransformers, is a new method of pre-training language representations which obtains state-of-the-art results on a wide array of Natural Language Processing (NLP) tasks. You can find the academic paper here: . BERT B E R T https://arxiv.org/abs/1810.04805 BERT has two stages: and . Pre-training fine-tuning is fairly expensive (four days on 4 to 16 Cloud TPUs), but is a one-time procedure. BERT has released a number of pre-trained models. Most NLP researchers will never need to pre-train their own model from scratch. Pre-training is inexpensive. One can replicate all the results given in the paper, in at most 1 hour on a single Cloud TPU, or a few hours on a GPU. For example, SQuAD can be trained in around 30 minutes on a single Cloud TPU to achieve a Dev F1 score of 91.0%. Fine-tuning So our first step is to Clone the BERT github repository, below is the way by which you can clone the repo from github. Now get inside the Bert repo using “ “ command cd !git clone https: //github.com/google-research/bert.git cd bert 3) Download the BERT PRETRAINED MODEL BERT Pretrained Model List : BERT-Large, Uncased (Whole Word Masking): 24-layer, 1024-hidden, 16-heads, 340M parameters BERT-Large, Cased (Whole Word Masking): 24-layer, 1024-hidden, 16-heads, 340M parameters BERT-Base, Uncased: 12-layer, 768-hidden, 12-heads, 110M parameters BERT-Large, Uncased: 24-layer, 1024-hidden, 16-heads, 340M parameters BERT-Base, Cased: 12-layer, 768-hidden, 12-heads, 110M parameters BERT-Large, Cased: 24-layer, 1024-hidden, 16-heads, 340M parameters BERT-Base, Multilingual Cased (New, recommended): 104 languages, 12-layer, 768-hidden, 12-heads, 110M parameters BERT-Base, Multilingual Uncased (Orig, not recommended): 102 languages, 12-layer, 768-hidden, 12-heads, 110M parameters BERT-Base, Chinese: Chinese Simplified and Traditional, 12-layer, 768-hidden, 12-heads, 110M parameters BERT has released and models, that have uncased and cased version. Uncased means that the text is converted to lowercase before performing Workpiece tokenization, e.g., John Smith becomes john smith, on the other hand, cased means that the true case and accent markers are preserved. BERT-Base BERT-Large When using a cased model, make sure to pass -do_lower=False at the time of training. You can download any model of your choice. We have used the BERT-Large-Uncased Model. !wget https: //storage.googleapis.com/bert_models/2018_10_18/uncased_L-24_H-1024_A-16.zip # Unzip the pretrained model
!unzip uncased_L _H _A zip -24 -1024 -16. 4) Download the SQUAD2.0 Dataset For the Question Answering task, we will be using SQuAD2.0 Dataset. (Stanford Question Answering Dataset) is a reading comprehension dataset, consisting of questions posed by crowdworkers on a set of Wikipedia articles, where the answer to every question is a segment of text, or span, from the corresponding reading passage, or the question might be unanswerable. SQuAD SQuAD2.0 combines the 100,000+ questions in SQuAD1.1 with over 50,000 new, unanswerable questions written adversarially by crowdworkers to look similar to answerable ones. You can download the dataset from SQUAD site https://rajpurkar.github.io/SQuAD-explorer/ #Download the SQUAD train and dev dataset
!wget https: !wget https: //rajpurkar.github.io/SQuAD-explorer/dataset/train-v2.0.json //rajpurkar.github.io/SQuAD-explorer/dataset/dev-v2.0.json 5) Set up your TPU environment Verify that you are connected to a TPU device You will get know your TPU Address which is used at the time of fine-tuning Perform Google Authentication to access your bucket Upload your credentials to TPU to access your GCS bucket Using code below you can do the above mentioned 4 points: datetime json os pprint random string sys tensorflow tf os.environ, TPU_ADDRESS = + os.environ[ ]
print( , TPU_ADDRESS) google.colab auth
auth.authenticate_user() tf.Session(TPU_ADDRESS) session:
  print( )
  pprint.pprint(session.list_devices()) open( , ) f:
    auth_info = json.load(f)
  tf.contrib.cloud.configure_gcs(session, credentials=auth_info) import import import import import import import import as assert 'COLAB_TPU_ADDR' in 'ERROR: Not connected to a TPU runtime; please see the first cell in this notebook for instructions!' 'grpc://' 'COLAB_TPU_ADDR' 'TPU address is => ' from import with as 'TPU devices:' # Upload credentials to TPU. with '/content/adc.json' 'r' as # Now credentials are set for all future sessions on this TPU. 6) Create an output directory : You will need a GCP (Google Compute Engine) account and a GCS (Google Cloud Storage) bucket to run the colab file. Please follow the Google Cloud for how to create a GCP account and GCS bucket. You have $300 free credit to start with any GCP product, learn more about it at . Prerequisite https://cloud.google.com/tpu/docs/setup-gcp-account You can create your GCS bucket from here . http://console.cloud.google.com/storage As we are using the Cloud TPU, we need to store the pre-trained model and the output directory in the Google Cloud Storage. If you are not storing it on Bucket you may face the following error : ERROR:tensorflow: recorded training_loop: From /job:worker/replica: /task: :
Unsuccessful TensorSliceReader : Failed to get matching files on uncased_L-24_H-1024_A-16/bert_model.ckpt: Unimplemented: File system scheme '[local]' not implemented (file: 'uncased_L-24_H-1024_A-16/bert_model.ckpt')
[[node checkpoint_initializer_14 (defined at /usr/local/lib/python3.6/dist-packages/tensorflow_core/python/framework/ops.py:1748) ]] Error from 0 0 constructor You will get your fine_tuned model in the Google cloud storage bucket after completion of training. For that, you need to provide your BUCKET name and OUPUT DIRECTORY name. BUCKET = BUCKET, output_dir_name = BUCKET_NAME = .format(BUCKET)
OUTPUT_DIR = .format(BUCKET,output_dir_name)
tf.gfile.MakeDirs(OUTPUT_DIR)
print( .format(OUTPUT_DIR)) 'bertnlpdemo' #@param {type:"string"} assert '*** Must specify an existing GCS bucket name ***' 'bert_output' #@param {type:"string"} 'gs://{}' 'gs://{}/{}' '***** Model output directory: {} *****' 7) Move Pretrained Model to GCS Bucket Need to move Pre-trained Model at GCS (Google Cloud Storage) bucket, as Local File System is not Supported on TPU. If you don’t move your pre-trained model to TPU you may face the error. The command allows you to move data between your local file system and the cloud, move data within the cloud, and move data between cloud storage providers. gsutil mv !gsutil mv /content/bert/uncased_L _H _A $BUCKET_NAME -24 -1024 -16 8) Training Below is the command to run the training. To run the training on TPU you need to make sure about below Hyperparameter, that is tpu must be true and provide the tpu_address that we have found above. -use_tpu=True -tpu_name=YOUR_TPU_ADDRESS !python run_squad.py \
  --vocab_file=$BUCKET_NAME/uncased_L _H _A /vocab.txt \
  --bert_config_file=$BUCKET_NAME/uncased_L _H _A /bert_config.json \
  --init_checkpoint=$BUCKET_NAME/uncased_L _H _A /bert_model.ckpt \
  --do_train= \
  --train_file=train-v2 .json \
  --do_predict= \
  --predict_file=dev-v2 .json \
  --train_batch_size= \
  --learning_rate= \
  --num_train_epochs= \
  --use_tpu= \
  --tpu_name=grpc:// : \
  --max_seq_length= \
  --doc_stride= \
  --version_2_with_negative= \
  --output_dir=$OUTPUT_DIR -24 -1024 -16 -24 -1024 -16 -24 -1024 -16 True .0 True .0 24 3e-5 2.0 True 10.1 .118 .82 8470 384 128 True Create Testing File We are creating input_file.json as a blank JSON file and then add the data in the file in the SQuAD dataset format. is used to create a file - is used to write a file in the colab - touch %%writefile You can pass your own questions and context in the below file. !touch input_file.json %%writefile input_file.json
{ : , : [
        { : , : [
                { : [
                        { : , : , : },
                        { : , : , : },
                        { : , : , : }
                    ], : }
            ]
        }
    ]
} "version" "v2.0" "data" "title" "your_title" "paragraphs" "qas" "question" "Who is current CEO?" "id" "56ddde6b9a695914005b9628" "is_impossible" "" "question" "Who founded google?" "id" "56ddde6b9a695914005b9629" "is_impossible" "" "question" "when did IPO take place?" "id" "56ddde6b9a695914005b962a" "is_impossible" "" "context" "Google was founded in 1998 by Larry Page and Sergey Brin while they were Ph.D. students at Stanford University in California. Together they own about 14 percent of its shares and control 56 percent of the stockholder voting power through supervoting stock. They incorporated Google as a privately held company on September 4, 1998. An initial public offering (IPO) took place on August 19, 2004, and Google moved to its headquarters in Mountain View, California, nicknamed the Googleplex. In August 2015, Google announced plans to reorganize its various interests as a conglomerate called Alphabet Inc. Google is Alphabet's leading subsidiary and will continue to be the umbrella company for Alphabet's Internet interests. Sundar Pichai was appointed CEO of Google, replacing Larry Page who became the CEO of Alphabet." Prediction Below is the command to perform your own custom prediction, that is you can change the input_file.json by providing your paragraph and questions after then execute the below command. !python run_squad.py \
  --vocab_file=$BUCKET_NAME/uncased_L _H _A /vocab.txt \
  --bert_config_file=$BUCKET_NAME/uncased_L _H _A /bert_config.json \
  --init_checkpoint=$OUTPUT_DIR/model.ckpt \
  --do_train= \
  --max_query_length= \
  --do_predict= \
  --predict_file=input_file.json \
  --predict_batch_size= \
  --n_best_size= \
  --max_seq_length= \
  --doc_stride= \
  --output_dir=output/ -24 -1024 -16 -24 -1024 -16 -10859 False 30 True 8 3 384 128 To make it easier for you, we have already created a Colab file which you can copy in your Google Drive and execute the commands. You can access the colab file at: . Question Answering System using BERT + SQuAD on Colab TPU If you have any further questions or doubt then please feel free to post them in comments. We’ll get back to you. Previously published at https://www.pragnakalp.com/nlp-tutorial-setup-question-answering-system-bert-squad-colab-tpu/ References Our case study Question Answering System in Python using BERT NLP BERT based Question and Answering system demo

NLP Tutorial: Creating Question Answering System using BERT + SQuAD on Colab TPU

aims to solve the problem of in a multi-party audio recording. Speaker Diarization “Who Spoke When” At , ML team is building an automated quality assurance engine for . During the quality check phase, a calling representative’s performance is scored on various quality parameters, such as: Squad SquadVoice Whether the representative was rushing on the call or not, Whether (s)he was rude to the lead at any time or not, Whether (s)he used the proper language, and many more. Such quality checks for a call recording are guided by “ ” was spoken by the calling representative and “ ” the same content was spoken. Thus, we need to ensure that the call recording portions where the agent spoke are separated from the portions where the customer or lead spoke. This, was a challenging problem, as Squad’s cloud telephony partner for India, records calls only in monaural format. Monaural format stores both parties audio on a single channel as opposed to stereophonic format, where audio of caller would be stored on one channel and that of callee would be written on a different channel. Thus, as a prerequisite for the quality checks, a speaker diarization system was required. However, we had a more focused problem, since the number of speakers for our use case was fixed at two. what how How to solve Speaker Diarization? The problem of speaker diarization is quite complex. To be honest, it is the toughest problem that I have worked on till date. This solution utilizes both supervised and unsupervised Machine Learning techniques. Also, it relies on a combination of both recent and conventional models. So let’s get started. Machine Learning Deep Learning Agglomerative clustering Our speaker diarization solution is derived from the combination of , speaker recognition architecture used by the work: ”, and speaker change detection algorithm explained in this work: ”. One shot learning “VoxCeleb: a large-scale speaker identification dataset “TristouNet: Triplet loss for speaker turn embedding Speaker differentiation pipeline We utilized three VGG-M based CNN architectures [with shared weights] to be trained using triplet loss optimization paradigm in order to learn (dis)similarity between audio recordings of same/ (different) speakers. We call this speaker differentiation pipeline. To train a speaker differentiation pipeline, on each iteration of model training, a triplet of audio recordings is generated, represented by <q_sound, p_sound, n_sound>. Here q_sound represents anchor audio snippet, p_sound represents some other audio snippet of the same speaker and n_sound represents an audio snippet of a different speaker. Then, magnitude spectrogram for each audio in the triplet is computed and pushed to VGG-M based model described above. All three CNNs share the same weights throughout the optimization process. Lastly, embeddings from the last layer of the network represented by <q_vec, p_vec, n_vec>are pushed to the hinge loss layer defined as: Loss function optimized for speaker differentiation In above equation D(x, y) represents the Euclidean distance between two embedding vectors and “g” margin constant in the hinge loss function, is kept equal to 1. Speaker differentiation pipeline visualization in this step: Few gotchas to be careful of Ensure the input magnitude spectrograms are mean and variance normalized along the time axis. Intuitively, each human would have few frequencies which would be more prominent in their voice, which contributes to the unique identity of sound produced by their vocal chords. This would ensure that energies corresponding to a frequency bin for different phonemes pronounced over a short time frame are smoothened. Use activation function in the last fully connected layer, instead of Relu, if you use Euclidean distance as the distance metric, this is because square root (x) is non-differentiable at 0. Or optimize squared Euclidean distance if you wish to use Relu. softplus Base CNN model trained for speaker differentiation One of the key benefits of using this model is that, it will give the same dimension embedding vectors for any length of audio input to the model. However, since the model was trained using batch gradient descent hence during training phase we fixed the audio recordings length at 3s. The dataset for Training Speaker Differentiation Pipeline: To train the pipeline, we utilized audio recordings from IVR mission calls that were placed through SquadVoice. We took call recordings of 100 speakers with 12.06 mins of audio per speaker on average, while each IVR recording being 2.41 mins on average. Additionally, we performed preprocessing on these IVR recordings, like the removal of IVR voice and trimming silence from the end of call recording using . sox Next, we computed random non-overlapping audio snippets of length between (4.5,s 12.5s), from each audio recording. Thus, generating a total dataset of 8586 audio snippets, Lastly, we took random 3s crop of audio recordings while passing them as input to the neural net on each iteration and ensured that both train and validation set audio snippets had non-overlapping speaker distribution. That is, there is no speaker whose audio snippet belonged to both train and validation split. Speaker Change Detection We applied the learnings from TristouNet work to find speaker change points using the model trained above. We maintained 2 sliding window pairs of left and right sliding windows having duration 3s and moved them at a very small stride of 0.1s. Then, for each left and right window pair, obtained their corresponding magnitude spectrograms and ran them through the CNN trained in step (1) to obtain embeddings for each left and right window pair. Afterward, we computed the Euclidean distance between both left and right embeddings, if the distance exceeds a particular threshold then that would represent a speaker change point. Example, for a 30s audio recording, we would have the following workflow: Workflow for speaker change detection Agglomerative clustering for grouping the audio snippets We broke down the audio as per speaker change points. I.e. if there were 5 speaker change points detected then we split the audio into 6 audio recordings, calculated magnitude spectrograms for the obtained snippets and then computed speaker embeddings using the CNN trained earlier. Later ran on top of these embeddings to group together audio snippets by their speaker. Agglomerative clustering Experiment Results and Evaluation This way the complete diarization pipeline was built. When tested on the dataset of 100 call recordings. It had an average of 12.23%, which is quite great as per industry standards. Additionally, the pipeline also performs well in presence of background noise. As examples, have added few sample diarized audio calls with their original versions. The attached recordings are mock calls following actual customer flow. diarization error rate [confusion part] +---------------+-------------+-------------+---------------+| original call | Calling Rep |    Lead     | Confusion DER |+---------------+-------------+-------------+---------------+| | | |       5.5%    || | | |       6.44%   || | | |       6.26%   |+---------------+-------------+-------------+---------------+ call_1 call_1_rep call_1_lead call_2 call_2_rep call_2_lead call_3 call_3_rep call_3_lead Acknowledgment I would like to thank , for discussions and reviews of the pipeline design and results of speaker diarization, along with and for design discussions and PR reviews while moving this pipeline to SquadAI codebase. Also would like to thank Operations team for their vital inputs. Sanchit Aggarwal Pragya Jaiswal Ved Vasu Sharma Squad Voice

Speaker Diarization — The Squad Way

This is Part IV in a series on how Automated Insights revamped its internal organizational structure and processes. Start at the beginning with Part I . In , I went into some detail about how we structured Squads for the whole company. In this article I’m going to talk about our Agile-inspired workflow. Part III As a guiding principle, we used the to provide a base to build from. Our version goes like this: Scientific Method Identify a problem using data Propose a way to improve a Squad metric related to the problem (hypothesis) Build a quick solution (experiment) Check the metric to evaluate impact Iterate and improve A Culture of Experimentation is Key The Scientific Method is based on the idea of using experiments to tests hypotheses. Finding answers through experimentation is something I’ve pushed at Automated Insights from the very beginning. I often talk about how Automated Insights is one big experiment. Because of my pessimism about traditional corporate culture, we experiment with virtually everything we do. If it works, we’ll keep it. If it doesn’t, we move on to the next experiment. When we adopted the Spotify Squad model for our engineering organization, it was an experiment. It turned out to be a successful one. We stressed with our Squads that experiments are the most important aspect of what we do. Build quick solutions to validate a hypothesis, then iterate, rebuild, or move on based on the results. A related concept that I’ve stressed with the Squads is “small over big.” I’d rather have small teams over big teams. Small meetings over big meetings. Small processes over big processes. Small solutions over big solutions. It’s a simple, almost trivial, heuristic, but it’s one that I find works well most of the time. Especially when you are trying to scale, small over big is almost always a better answer. Our Sprint Cycle We chose two weeks for the length of our sprints. Why two weeks? It was long enough to split up work and make good progress, but short enough to allow Squads to stay agile and iterate. Some organizations use longer or shorter sprint cycles. Two weeks has worked for us. I can’t imagine having it any shorter because that means the planning/demo meetings are taking a larger percentage of the sprint (one out of five days instead of one out of ten days). We’ve discussed the possibility of extending it to three weeks, which would make for a good experiment at some point. For our team members that were not previously familiar with Agile (especially those in non-engineering groups like HR, Finance, etc.) we had to stress that not everything has to be completed in two weeks. A “project” could take two months and that’s totally acceptable, but you’d be hard pressed to convince me that a two month project can’t be broken up into several one or two week chunks of work. And if it can’t, then how can we possibly track progress and estimate with any certainty? It’s important to understand that two weeks is simply a progress measure for the work, it’s not a rigid timeframe that we must make every project or feature timeline conform to. That said, I often hear someone quoting that it will take “two more sprints worth of work” or “a sprint and a half”. It’s pretty natural to think of how long something will take in terms of the sprint duration. However, it is an imprecise measure (as is often the case with software development), which is why I’m less keen on having longer sprint cycles. Two sprints could mean 4 weeks or maybe it means 3.75 weeks or even 3.5 weeks. If we had three week sprints, it would be even more imprecise to say something is a “sprint” worth of work. With a two week sprint duration, that means we get (re)organized every two weeks. Our big planning day is on Monday (i.e., every other Monday). On those Mondays we have two key activities: company-wide demo meeting and individual Squad planning sessions. Company-wide Demo Meeting Prior to implementing Squads, I scheduled one-on-ones with everyone in the company. That meant roughly fifty, thirty minute one-on-ones over a couple weeks. This was just a one time event so I could get a quick pulse on the company. The most consistent issue I heard was that as we got bigger, it was harder to keep track of what everyone was working on. And eventually they’d see something new going live and was surprised because they didn’t know someone was working on it. Our bi-weekly demo meeting solves that issue. Now, every other Monday, I require everyone get up and spend up to two minutes either showing off something they worked on over the last sprint (i.e. demoing) or giving an update on the things they did. It’s the most important meeting we have at the company. Initially, we scheduled this meeting for three hours. With 50 people giving two minute updates each plus setup time, we thought it could go three hours. Fortunately, they’ve gone two hours consistently, especially as we gotten better at the hand-offs between each person. We’ve also moved from everyone having presentations on their computer, to a central repository (intranet) of slides and docs that we can show from a single computer. Instead of developers having a terminal window open to show off their work, we use animated gifs that highlight the key code or features they want to talk about. There are a few things to mention about the Demo meeting. I’m not a fan of meetings in general, so it’s a little ironic that I condone and even advocate everyone attending the bi-weekly Demo meeting. I do not everyone attend the Demo meeting, but I encourage it. We record the full meeting so it can be watched later if you can’t attend the whole thing. I do require that everyone be present to do . Feel free to pop in and out, but just don’t miss your demo slot. require their demo Why do I want everyone to demo? There are a bunch of good reasons and virtually no bad ones. First, at a company like Automated Insights, we need everyone to be able to communicate their ideas and thoughts clearly. If you can’t get up in front of your friends and colleagues for two minutes and talk about what you did for the last two weeks, you are likely not going to be a very effective employee. I’m not saying it will be easy for everyone due to fears of public speaking, but we have a round of applause after each demo presentation and are cutting jokes throughout the two hours, so it’s a pretty low stress environment. It also provides a little extra incentive to put your best foot forward during each sprint. You don’t want to get up in front of the whole company and try to bluff your way through your update. It’s much easier if you can show something interesting. Early on some people would come to me and say “But I don’t have anything to show”. That’s fine, just give a quick summary on the key features, bugs, processes, servers, meetings, interactions, discussions, trips, whatever that you were involved with. Two weeks is long enough that everyone should have done something that is worth mentioning even if that was “refactored the users page” or “evaluated our HRMS options.” I much prefer employees show us something they worked on, but it’s ok to simply tell us if it is not possible to show it. But I strongly encourage showing it. An unexpected side affect of the Demo meeting is it has turned into a forum for employees to show photos from vacations or fun excursions they’ve gone on. It’s not uncommon for someone to take a minute to show pictures from a recent trip or something funny their kid did over the weekend or sharing news about a new wedding engagement. In summary, our Demo meeting is a 2 hour all-hands meeting that’s optional but strongly encouraged where each employee spends up to 2 minutes updating the company on their recent work and personal events. Not a single person has come to me in the last year to say they feel disconnected from other parts of the company. Sprint Planning Meetings After the Demo meeting, the Squads are encouraged to get together to start planning out the next sprint. The sprint planning meetings can last anywhere from 30–60 minutes. Early on when we were just starting, some needed 90 minutes. We aren’t rigid about how Sprint Planning meetings are run. Autonomy is a key feature of the Squad model, so we stay away from dictating much about how Squads do their thing. We give them what we consider are best practices, but leave it up to them to implement. The goal for the planning meetings is to leave with a plan for the next two weeks. The Tribe Leader for the Squad generally attends so she can stay in the loop with the Squad and nudge a Squad in a particular direction if needed. The basic template agenda for a Sprint Planning meeting goes like this: review the Squad’s metrics and general trends Metrics: discuss in-progress or completed experiments and any impact on the metrics Experiment review: prioritize next problems to work on Problem ranking: as a group, develop solutions to those problems Feature planning: discuss roadblocks and ways to improve the sprint process Retro: Most Squads use Trello to keep track of their work. Here again, we have a basic template we use for each Squad’s Trello board, but they can deviate from it if needed. Sample Squad Trello Board Other Squad Meetings We recommend that Squads have one or two a week to check-in on progress. These are quick meetings that can help ensure everyone stays in sync and is knowledgeable about any roadblocks. Some Squad Standups are in-person and some are done virtually via Slack. We’ve seen a lot of experimentation with different formats. In my experience, Standups rarely last without tweaking periodically. They can get a little mundane (“I have nothing to update since last Standup”) so often Squads are trying to tweak the format to make them more effective. Standup meetings Another type of Squad meeting that is pretty common is a design session for the Engineering-oriented Squads or just general brainstorming meetings for non-engineering Squads. These are necessary when you’ve decided to build a new feature or implement a new “thing” during the Sprint Planning meeting, but need to spend more time specing it out. Sometimes the results of a design session will require a second, briefer Squad Planning meeting once additional details are known about how much effort a certain feature will take to implement. Squad Leader Meeting From the beginning, we had an hour long Squad Leader meeting the Tuesday following the Demo meeting (so every other Tuesday). Initially, the purpose of the meeting was to answer questions Squad Leaders had about the new Squad model as well as get their feedback about what was working or not working. Once everyone was comfortable with the Squad model, the agenda for the meeting settled into this: changes and trends; issues and difficulties in moving the numbers Update on metrics: : changes in how the squad is impacting the system; potential changes to metrics or measurement Impact on the System : this sprint’s activities that may impact other squads; future activities that may impact or require collaboration from other squads Intra-Squad communication : discuss any resource gaps or surpluses within Squads as well as any upcoming Squad resource changes Proposed process/squad changes : how to have hard conversations; best practices for managing performance; etc. Special topics As we’ve reached a stasis, the Squad Leader meeting has turned into more of a status meeting ( ), so we’ve discussed the possibility of moving the meeting to a wiki on the intranet that can be done virtually instead of in-person. which isn’t necessarily a good thing Tribe Leader Meeting Similar to the Squad Leader meeting, the Tribe Leader meeting has been a forum for the five Tribe leaders to get together every couple weeks. Unlike the Squad Leader meeting, this meeting has been more sporadic in nature and we went several months without having it at all. Our Tribe Leaders are already pretty connected with each other so a separate meeting isn’t usually necessarily. Occasionally, if we start to feel disconnected or if there are particular issues we need the Tribe Leaders to discuss, we’ll schedule an ad-hoc meeting. Guild Meetings As I described in , we have cross-Squad interest groups called Guilds. There is a Developer Guild, a Designers Guild, etc. Each Guild holds a meeting periodically to get the Guild members together to discuss relevant topics. We haven’t put in place any guidelines for how Guild meetings should work. They are self-organizing and self-governing. Part III The Developer Guild has been our most successful Guild to date. They meet every two weeks and keep a running list of prioritized topics to discuss. Also, they have one or more people do quick Lightning Talks where they show something cool they’ve been working on or a new product/API/tool they’ve been using. The Developer Guild Leader has been great about sending out notes from the meeting to the other Squad Leaders to keep them in the loop. Uniformity, but with Autonomy As the CEO with a traditional organization model, keeping track of how all the different groups in the company operated was challenging, especially as we grew. It was also difficult to know how everyone was performing and contributing to the company goals. By having everyone adopt the same basic operating model and cadence, it makes it a lot easier to understand what any one person or group is working on and when work is expected to be done. While we have uniformity in terms of the basic structure and operating model, within each Tribe and Squad there is autonomy to tweak the day-to-day processes as needed. This has provided the structure the leadership teams needs to run the company, but the freedom our employees need to work the way that best suits their team. Next up in Part V, I’m going to talk about how the traditional role of “manager” is broken and how we split it apart inside of Ai. ❤️ If you like this article, please let me know by clicking the button below!

Inside Ai Part IV: Getting Work Done with Squads

Aligning our organizational structure with the customer journey This is Part III in a series on how Automated Insights revamped its internal organizational structure. Start at the beginning with Part I . In the , I described how we implemented a version of in our Engineering organization. As we fully committed to transitioning from a services company to a product company, it was time to rethink the organizational structure for the rest of the company too. Previously, we were functionally aligned (i.e. Engineering, Sales, Marketing, etc.) as is common in companies large and small. last article Spotify’s Tribes and Squads Traditional Org Chart A big challenge with that model is you can lose sight of the customer. One of is . Employees, especially engineers down in the weeds of creating features and fixing bugs, can sometimes forget the real reason we do what we do. Our value is simply a reminder that Automated Insights is able to stay in business because of customers. It doesn’t mean “the customer is always right” or anything like that, but we can’t simply create features only because think it is a good idea or fun to do. It must also be something that a customer ultimately wants. our company values Think Like a Customer Think Like a Customer we With that context, the functional organization model doesn’t align well with the . If you want to improve or change some aspect of the customer experience, it is often challenging to assign end-to-end ownership for that change. Let’s say we have a goal of making the free trial process more successful. Does Product Management own that since they spec out the features? Or Engineering because they‘ll build the features? Or should it be Sales because they have the most direct interaction with customers doing trials? customer journey It was in a meeting with our Chief Revenue Officer, Adam Smith, and VP of Product Management, Adam Long, that we discussed aligning our Squads with the customer journey. That way, responsibility for key features and processes (from the customer perspective) would have clear ownership. Much of the rest of this article came from brainstorming sessions the three of us had prior to implementing Squads company-wide. Aligning Squads with the Customer Journey First, we needed to define the customer journey for our Wordsmith product. This was a good exercise as we had never thought about the end-to-end process that a customer goes through when interacting with us as a company. A simplified version of that is as follows: First, you about Wordsmith, then you about its capabilities, then you want to it out, and you do that by an NLG template, then you’ll want to refine ( ) that template until you are ready to your content. hear learn try writing edit publish The initial Wordsmith customer journey map Next, we mapped each of those functions to a Squad and thought about what was left over from our original organizational chart. The customer journey doesn’t map to every job you need to do inside a company (e.g. running payroll or keeping servers running). Some Squads aren’t aligned with a defined customer interaction, but that’s ok. We were coming from a model where there was virtually no alignment with the customer journey, so this was at least a step in the right direction. What is a Squad? A Squad is another name for a team. Nothing revolutionary there, but in the Spotify model, Squads are a particular type of team. They should be small, highly aligned, and loosely coupled. Each Squad should have the following attributes: : how will they help customers at a particular stage or enable other Squads to do their work Clear mission : self-organizing team that can attack their stage’s problems Clear autonomy : optimize the 1–3 metrics that measure performance for the Squad at its stage in the customer journey Clear accountability each person in the organization should be on only one squad (we had an issue where our best employees were spread too thin) Focused: composed of people from different disciplines (dev, design, QA, support, etc.) Cross-functional: Ideally, each Squad is completely autonomous and has all of the resources it needs to accomplish its work, but in a 50 person company that’s just not practical. It would be hard to implement this model in companies that are less than 30-40 people. You just don’t have enough employees to fill out all of the Squads that map to the customer journey. And one thing we learned is single person Squads do not work well. In the Ai model, each Squad has just two defined roles: Squad Leader and Squad Member. The Squad Leader is the one that is ultimately responsible for the performance of the Squad (and therefore its members). I’ll go into a lot more detail about the role of Squad Leaders in Part V. A Tribe is a collection of similarly aligned Squads. Each Tribe has a Tribe Leader that works closely with the Squad Leaders in the Tribe. The Tribe Leader helps ensure Squads within the Tribe don’t conflict with one another. The Tribe Leader is also responsible for helping define the mission and metrics for each of their Squads in collaboration with the Squad Leaders. Ai Tribes and Squads We came up with sixteen Squads within five Tribes that covered all aspects of our business: The Attract Tribe consists of our old Sales and Marketing functions. We inserted a Buy Squad after Learn and before Try even though customers generally don’t Buy before they Try. This was the first of a few compromises we had to make so the model would work for us. The Buy Squad in this context is our old Sales team. Since the Attract Tribe consisted of Marketing folks and the Build Tribe consisted mostly of Engineering and Product Management folks, it made more sense to have Buy be a part of the Attract Tribe (whose Tribe Leader was our CRO). While not entirely consistent with the customer journey, we decided early on that we would not be held slave to a rigid structure. We used the customer journey as more of a guide than a strict rule. Another key feature of this model is that we wanted it to be fluid over time. We recognized we were trying something pretty different, so I wanted everyone to be comfortable with iterating on the model regularly. One of the major failings of the traditional organization is its inflexibility to change. I wanted to fix that. For the Squads in the first three Tribes (that mapped with the customer journey), we named them after verbs that were associated with the action the customer was performing. This was one way to try and break away from the traditional org names (Sales becomes Buy, Marketing becomes Hear and Learn, etc.). The Build Tribe was made up of engineers and product managers building the core Wordsmith product. The Scale Tribe consisted of our managed services team, which had a mix of data scientists and developers, as well as our account management team. That still left several functions that were important but not represented in the customer journey. We created an Enable Tribe whose main purpose was to help the other Squads get work done. It was made up of our Research (Labs), DevOps, Security, and Analytics teams. Finally, that left Finance, HR and Legal, which we created an Ops Tribe for. The main benefit of having Ops included in our Squad model has more to do with how we get work done than necessarily aligning with the customer journey, which I’ll describe in the next article. Guilds While our Squads covered most of the functions we could think of, there were still a few issues. For example, with developers spread out across at least nine of the sixteen Squads, how would we ever keep them on the same page? Likewise, when it comes to design, how do we make sure marketing folks in the Attract Tribe are using the same style guidelines as designers in the Build Tribe? The answer is Guilds. A Guild is simply a self-organized interest group. Out of the gate we did not specify any Guilds. We wanted to see what would form organically and sure enough, they did. Guilds have the additional feature of providing leadership opportunities for people that may not be a Squad Leader yet. If you feel ready to show your leadership potential, form a Guild. Guilds typically set standards for how to do a particular role across the company as well as share lessons learned. Spotify also described the concept of a , which is similar to a Guild, but contained within a particular Tribe. That may be necessary when you are their size, but we haven’t found a need to carve out functions within a Tribe that aren’t relevant across all Tribes. Chapter Cross-functional Guild Summary We covered a lot of ground in this article including why we adopted the Spotify Tribes and Squads model and how we mapped it to the Wordsmith customer journey. Here are a few key points about Squads: Main benefits of this model include autonomy, agility and focus. One Squad per person — more than one Squad per person indicates a resource gap. Being on one Squad helps drive accountability and focus. Each Squad has 1–3 metrics they are held accountable for. The Squad Leader is accountable for all work produced by the Squad. The Tribe Leader is accountable for all work produced by their Squads. The Squad structure will be fluid over time. Over the past year, we’ve seen many benefits from this model. While no organizational model is a panacea that will solve all your problems, we found that the Squad approach offered: Greater accountability Better execution Clear focus areas Cross-functional alignment That does it for this article. Next up in Part IV, I’m going to talk about how we get work done with Squads. If you are familiar with , it won’t be anything new — except we make every Squad use Agile including HR! Agile ❤️ If you like this article, please let me know by clicking the button below.

Inside Ai Part III: Squadifying the Company

A new initiative has started in . We call it . Javascript Israel Goodness Squad The main objective is to meet once a month for a coding event which is all about fun, learn and contribution. is structured as a hackathon. First of all, there’s beer. Then, a list of 4–6 projects are presented. We pick these projects before the event, and make sure each project has an owner, a project lead. Each project lead makes sure the project is accessible for newcomers. This means it should be clear what contribution the project is expecting. Either there’s a list of , or just unassigned issues, documentation, unit tests, user feedback etc. A project can define for itself what help is needed from the community. In addition, it should be super easy to setup the environment for local development. The event good first bugs Our first event was 2 weeks ago, on January 17. We were kindly hosted by the amazing at their offices in Tel Aviv port. There were 25 participants and 6 projects: Wix MobX by Mozilla Debugger.html Angular CLI Angular Material Redash write.io What I absolutely love about this list is the diversity. Some projects are big community projects, some are backed by organizations, some maintained by small startups and some are just initiatives of individuals. We had at least one of everything. This is how it looked: What can a developer do in 3 hours? The most prominent drawback people had when hearing about this was — “Are 3 hours really enough time to get into a new project?” The answer is yes. The added value is that each project’s mentor is able to get people rapidly into an existing project’s process and code. It was a very positive experience for people to be able to spot a bug and after 2 minutes of consultation with the project mentor do the triage and spot the relevant line of code. As one developer put it — “ you just saved me weeks of work”. Let’s back that up with data — here’s a list of issues and PR’s that Goodness Squad touched: https://github.com/search?utf8=%E2%9C%93&q=%22%23GoodnessSquad%22&type=Issues&ref=searchresults It’s not just about knowing lines of code. Process is also important. For example, Mozilla communicates in IRC. For a new contributor it’s sometimes intimidating to get on a channel and start spamming it with total noob questions. But in the event everyone opened up IRC and communicated directly with debugger.html maintainer, Jason. It was a real celebration. A response from the other side What turned this into a truly unprecedented event (for me, at least) was the response from the project owners themselves. MobX author tweeted back at us with cheers. Debugger.html lead wrote a direct thank you message to the contributors. Michel Weststrate Jason Laster This was received really well. Our slack channel was roaming with . People enjoy seeing impact and instant gratitude. We have reached win-win. party parrots FTF FTW It’s not only about contribution. It’s also about interaction and community. I strongly believe that we as developers cannot reduce our work to interface only via the machine. And Goodness Squad is a real opportunity to make programmers meet one another, discuss, have fun together and leverage the common ground which is our passion for creating in code. I do plan to make improvements to the program. Most of all, I want people to feel comfortable coming even if they feel they don’t have enough knowledge. It’s also a learning opportunity, and I’m sure everyone has something to contribute. Second, I do care about balancing women participation. In January we had zero women on board. For next event we already make sure to have women both participating and leading projects. Third, I would love to play music during the coding session. This might become a challenge to find something that works for everyone. I’ll keep you updated on that :) That’s is. February event (Wednesday 15th) is just around the corner and we’re working hard to make a great evening of coding, beer, fun and goodness. Peace out.

Goodness Squad

February last year I started doing referral sales on the platform. The idea of Squads is to connect remote, on-demand teams to clients that have a need for developing software products (web & app). I often ran into entrepreneurs or intrapreneurs wanting to build a great digital innovation. In most cases these people didn’t have the skills to design & build the required software. Some of them turned to hiring employees or working with freelancers. They lost a lot of time and money getting their development up to speed and ended up with flaky products. Squads.com When I came across Squads I was enthusiastic from the start. On this platform there are stable, high performing teams that can design & build software and are available on demand (24*7). It is the greatest solution for the issues that my fellow entrepreneurs face again and again. That’s when I joined Squads for business development and set-up a referral sales program. I use this referral program to earn a passive income through the platform. With the money earned, I started building my own Startup called SwipeGuide. SwipeGuide is a mobile, visual, step-by-step instruction app — ! check it out In this post I will share some tips & tricks that enable you to do the same. You can earn a good passive income or collect the budget to build your own tech startup. 1. Reach out to people that have software needs It’s clear that Squads focuses on software development decision makers. On the one side these are project owners of digital projects in companies. On the other side entrepreneurs building tech startups. Squads runs projects for both types of clients. In my network, I also have people in both categories, so I started reaching out to them at meetups and online. Everywhere I go I tell people about Squads. Not only because I want to get the referrals, but also because I am truly enthusiastic about the Squads approach. If you don’t believe it, don’t sell it. Especially not to people in your network. 2. Be creative with the referral link When you create an account on Squads.com you get a personal referral link. I use this one in my e-mail signature to point people to the platform for sign-up. I also make personalized messages in order to get people to sign-up. Within I have seen people using the link in their LinkedIn & profile as well. And I am sure there are more creative ways of using the referral link to get sign-ups. Every person that signs up through your referral link is visible in your account. Once they start booking capacity on Squads, you get the referral fee automatically added to your credits. Now I see the credits in my account build up week-by-week. This results in a passive income that I can either cash in or use directly to get my own software developed. StarterSquad Facebook This is how I added the referral link to my e-mail signature 3. Guide your clients Whenever a person signs up through your referral link, you need to give some guidance. They are, after all, on the platform through you, so there is a relationship of trust. You also want to help them get the most out of Squads. After signing up, I simply guide my referrals through a few simple steps getting started: : In Squads you can easily create a project by filling out the project briefing. This is a high level description of what you want to build. Although most of my referrals can complete this independently, I sometimes do a Skype call to help them through this process. I also inform some teams about the project that is coming though the Slack communication platform Squads is using. 1) Create a project : The next crucial step is selecting a team. Squads will suggest up to three teams based upon the project briefing. The smart algorithm in the platform normally does a good job, but I tend to have look as well. Team reputation & ratings give a lot of info, but I also have my personal favorites that I sometimes suggest to my contacts. In the end, the project is matched with a team and the show can start. 2) Select a team : Before the project can start money needs to be in the account. So my referrals have to buy some credits for the first week. This is easily done through the Squads payment options like paypal, creditcard, IDEAL and even bitcoin. I always highlight that there is a 100% satisfaction guarantee, so you keep control over your money. Budget can be allocated per week as all Squads teams work in weekly sprints. You set your first budget, discuss and prioritize the workflow, and your team starts working. By the end of the first week your contact gets a first demo, accepts the work and sets a new budget for the next week/sprint. Every time you’ll get a referral fee on the purchases, slowly building up your passive income. 3) Buying credits & start Squads.com (formerly known as Capacitor) short Explainer video 4. Use the earned credits for your projects After a while I earned a significant budget through Squads and was able to get an MVP built. That is when I kicked-off my SwipeGuide project. For SwipeGuide I also worked with a Squads team myself to do the design & software development for it. First, I worked with the Spotless team to design & build a prototype. Then I joined the Utrecht Inc accelerator with that product and pivoted from a B2C product to a B2B focus. At this point I chose to go for mobile first and switched to the Sirius team. The next release was built on my referral earnings combined with first payments of three launching customers. Squads provided an excellent springboard to get my kicked off. Startup 5. Next level referrals through growth hacks Initially, I only did basic stuff with my referral. In most cases I did personal selling. Also, I used my referral link in my e-mail signature. I created some blog posts that have my personal referral link as well. After studying the field of growth hacking, I now believe that opportunities exist to further improve my referral business. On Growth Hackers blogs and in books like ‘Traction’ you can find some interesting sales channels and growth hacks. These can be applied to referral sales on Squads.com as well. I’m convinced that I can create more referrals when I do some creative campaigns. This will be great as I have some new Startup ideas that I want to earn a budget for. So, if you see me send out messages regarding software development teams or Squads.com be sure to click the link, sign-up and start your project on Squads.com. By doing so you’ll help me build some great new software Startups that will benefit the world ;-) **Do you also want to earn through referrals?** Sign-up on Squads.com is how hackers start their afternoons. We’re a part of the family. We are now and happy to opportunities. Hacker Noon @AMI accepting submissions discuss advertising &sponsorship To learn more, , , or simply, read our about page like/message us on Facebook tweet/DM @HackerNoon. If you enjoyed this story, we recommend reading our and . Until next time, don’t take the realities of the world for granted! latest tech stories trending tech stories

How I built my Startup on Squads referral fees

Is a Green Build and Fixing Flaky Tests Your #1 Priority? I recently had a discussion with the founder of a CI delivery platform about the value of fixing flaky tests, and it led to the following observation/question on LinkedIn: “Over my career, I’ve been in maybe 50 software development teams as an IC. In my current role, I get exposed to 10,000+. If I picked a random person across my entire sample set, they wouldn’t even know where to start with ‘doing the right thing.’ I see way more disengaged engineers who have a ‘give me a ticket and I’ll give you code’ mindset. How would you go about rebooting that?” It’s a great question. I thought I’d write about how we’ve approached this in my last two companies. It’s easy to say that culture is the answer, but how do you build that culture? How do you get every member of the team to care about doing the right thing — and what is the right thing? So Many Priority #1s It often feels that everything is a top priority. But that, in itself, is part of the problem. Particularly in start-ups and scale-ups, Engineering and Product teams experience immense pressure to deliver. This pressure is often self-imposed - at most of my previous companies, stakeholders were generally willing to accept a temporary slowdown for a long-term increase in quality. Here’s the thing: achieving a stable environment with few bugs takes time, but once you’re there, you can actually deliver faster and more efficiently. With so much emphasis on features and delivery, it’s no wonder engineers feel torn between addressing quality and stability issues or pushing out new features. So, how do you approach this? While the question of priority is nuanced and will vary from company to company, here’s a list of priorities based on my recent experiences. 1. Build a Culture of Ownership and Accountability Building a culture of ownership takes time. Like any change, it requires repetition and reinforcement. Start immediately, as this will form the foundation for the following priorities. 2. Record, Prioritize, and Monitor Production Bugs and Defects It’s surprising how often teams fail to record and prioritize production issues. At UberCarshare, we had a successful product but no way to track bug count or severity. Once you record your production issues, you can start measuring and reporting them. This opens up meaningful conversations with stakeholders about the trade-offs between quality, stability, and new features. 3. Complete Root Cause Analysis (RCA) or Incident Reports for Significant Production Issues This complements the previous item and is a fundamental pillar of creating a culture of transparency, accountability, and autonomy. RCAs should be blameless: a celebration of transparency, learning, and self-improvement. When engineers feel comfortable writing RCA’s and see that transparency and accountability do not mean blame or punishment, they will feel a greater level of psychological safety. Tech debt is the root cause of many issues, and while we can’t fix everything immediately, completing RCAs and prioritizing follow-ups against other work is a major step toward prioritizing stability and quality over new features. 4. Establish a Continuous Green Build in CI I recognize there is an argument that this is a pre-requisite to any of the above but, like I said, it’s nuanced 🤷‍♂️, and a lot of this stuff can happen in parallel. In both of my previous companies, a flaky, unstable CI had become the norm, which is common in teams trying to scale quickly. You cannot overstate how expensive this is in terms of its impact on productivity, but when there is unclear ownership and accountability, it can be hard to know where to start in fixing it. Getting to a stable build can be a significant effort and may require dedicated resources. Our approach was as follows: Get the build consistently green and fast enough — exclude flaky tests as necessary! Address performance or timeout issues through sufficient resourcing and parallelization of tests.


Clarify and assign ownership of the excluded flaky tests — the squads owning them can decide whether they need to be fixed and reintroduced. Maintain zero tolerance for flaky tests: if tests become flaky, exclude them from the build until they are fixed 5. Monitoring System Health and Actually Caring About Your Logs It’s easy to set up error logging and reporting. However, getting your team to care about the logs and respond to them in real-time is far more challenging. Like performance and broken builds, system health needs to be a whole team effort and cannot be left to a DX, DevOps, or maintenance squad. System health relies on culture as much as anything else and once again requires a zero-tolerance approach. Agree on your tolerance thresholds with the team, ensure that alerts for exceeded thresholds are sent to a public channel, and set an expectation that no one in the team is allowed to walk by a problem. At Carshare, we established an expectation that no one should push an alert further into history by adding a new comment below without first ensuring someone was investigating the issue. Use These Principles to Drive a Culture of Autonomy and Accountability Document these principles and ways of working, workshop with the team on how best to implement them, share them with your product team and stakeholders, and then demand continuous improvement from your team. RCAs, error alerts, and broken builds should be attended to without prompting. Initially, you may need to prompt, nag, coach, and mentor your team. Above all, reassure them that fixing alerts, broken builds, or bugs really is their highest priority. Eventually, these issues will start to solve themselves, and productivity and output improvements will follow the noticeable improvement in quality and stability. Rinse and Repeat: The Path to Continuous Improvement The five priorities I’ve outlined are just the beginning. Continuous improvement should be ingrained in your process, and there will be many things to perfect: writing efficient tests
improving build speed
reducing handoffs
shifting quality left
improved collaboration with partners such as product and design teams
measuring quality, uptime, and cycle time
etc. Originally published at https://medium.com/vertical-slice/getting-software-engineers-to-do-the-right-thing-d42a0b923408 Is a Green Build and Fixing Flaky Tests Your #1 Priority? I recently had a discussion with the founder of a CI delivery platform about the value of fixing flaky tests, and it led to the following observation/question on LinkedIn: “Over my career, I’ve been in maybe 50 software development teams as an IC. In my current role, I get exposed to 10,000+. If I picked a random person across my entire sample set, they wouldn’t even know where to start with ‘doing the right thing.’ I see way more disengaged engineers who have a ‘give me a ticket and I’ll give you code’ mindset. How would you go about rebooting that?” “Over my career, I’ve been in maybe 50 software development teams as an IC. In my current role, I get exposed to 10,000+. If I picked a random person across my entire sample set, they wouldn’t even know where to start with ‘doing the right thing.’ I see way more disengaged engineers who have a ‘give me a ticket and I’ll give you code’ mindset. How would you go about rebooting that?” “Over my career, I’ve been in maybe 50 software development teams as an IC. In my current role, I get exposed to 10,000+. If I picked a random person across my entire sample set, they wouldn’t even know where to start with ‘doing the right thing.’ I see way more disengaged engineers who have a ‘give me a ticket and I’ll give you code’ mindset. How would you go about rebooting that?” It’s a great question. I thought I’d write about how we’ve approached this in my last two companies. It’s easy to say that culture is the answer, but how do you build that culture? How do you get every member of the team to care about doing the right thing — and what is the right thing ? right thing So Many Priority #1s It often feels that everything is a top priority. But that, in itself, is part of the problem. Particularly in start-ups and scale-ups, Engineering and Product teams experience immense pressure to deliver. This pressure is often self-imposed - at most of my previous companies, stakeholders were generally willing to accept a temporary slowdown for a long-term increase in quality. Here’s the thing: achieving a stable environment with few bugs takes time, but once you’re there, you can actually deliver faster and more efficiently. With so much emphasis on features and delivery, it’s no wonder engineers feel torn between addressing quality and stability issues or pushing out new features. So, how do you approach this? While the question of priority is nuanced and will vary from company to company, here’s a list of priorities based on my recent experiences. 1. Build a Culture of Ownership and Accountability Building a culture of ownership takes time. Like any change, it requires repetition and reinforcement. Start immediately, as this will form the foundation for the following priorities. 2. Record, Prioritize, and Monitor Production Bugs and Defects It’s surprising how often teams fail to record and prioritize production issues. At UberCarshare, we had a successful product but no way to track bug count or severity. Once you record your production issues, you can start measuring and reporting them. This opens up meaningful conversations with stakeholders about the trade-offs between quality, stability, and new features. 3. Complete Root Cause Analysis (RCA) or Incident Reports for Significant Production Issues This complements the previous item and is a fundamental pillar of creating a culture of transparency, accountability, and autonomy. RCAs should be blameless: a celebration of transparency, learning, and self-improvement. When engineers feel comfortable writing RCA’s and see that transparency and accountability do not mean blame or punishment, they will feel a greater level of psychological safety. Tech debt is the root cause of many issues, and while we can’t fix everything immediately, completing RCAs and prioritizing follow-ups against other work is a major step toward prioritizing stability and quality over new features. 4. Establish a Continuous Green Build in CI I recognize there is an argument that this is a pre-requisite to any of the above but, like I said, it’s nuanced 🤷‍♂️, and a lot of this stuff can happen in parallel. In both of my previous companies, a flaky, unstable CI had become the norm, which is common in teams trying to scale quickly. You cannot overstate how expensive this is in terms of its impact on productivity, but when there is unclear ownership and accountability, it can be hard to know where to start in fixing it. Getting to a stable build can be a significant effort and may require dedicated resources. Our approach was as follows: Get the build consistently green and fast enough — exclude flaky tests as necessary! Address performance or timeout issues through sufficient resourcing and parallelization of tests. Clarify and assign ownership of the excluded flaky tests — the squads owning them can decide whether they need to be fixed and reintroduced. Get the build consistently green and fast enough — exclude flaky tests as necessary! Address performance or timeout issues through sufficient resourcing and parallelization of tests. Get the build consistently green and fast enough — exclude flaky tests as necessary! Address performance or timeout issues through sufficient resourcing and parallelization of tests. Clarify and assign ownership of the excluded flaky tests — the squads owning them can decide whether they need to be fixed and reintroduced. Clarify and assign ownership of the excluded flaky tests — the squads owning them can decide whether they need to be fixed and reintroduced. Maintain zero tolerance for flaky tests: if tests become flaky, exclude them from the build until they are fixed Maintain zero tolerance for flaky tests: if tests become flaky, exclude them from the build until they are fixed 5. Monitoring System Health and Actually Caring About Your Logs It’s easy to set up error logging and reporting. However, getting your team to care about the logs and respond to them in real-time is far more challenging. Like performance and broken builds, system health needs to be a whole team effort and cannot be left to a DX, DevOps, or maintenance squad. System health relies on culture as much as anything else and once again requires a zero-tolerance approach. Agree on your tolerance thresholds with the team, ensure that alerts for exceeded thresholds are sent to a public channel, and set an expectation that no one in the team is allowed to walk by a problem. At Carshare, we established an expectation that no one should push an alert further into history by adding a new comment below without first ensuring someone was investigating the issue. Use These Principles to Drive a Culture of Autonomy and Accountability Document these principles and ways of working, workshop with the team on how best to implement them, share them with your product team and stakeholders, and then demand continuous improvement from your team. RCAs, error alerts, and broken builds should be attended to without prompting. Initially, you may need to prompt, nag, coach, and mentor your team. Above all, reassure them that fixing alerts, broken builds, or bugs really is their highest priority. Eventually, these issues will start to solve themselves, and productivity and output improvements will follow the noticeable improvement in quality and stability. is Rinse and Repeat: The Path to Continuous Improvement The five priorities I’ve outlined are just the beginning. Continuous improvement should be ingrained in your process, and there will be many things to perfect: writing efficient tests improving build speed reducing handoffs shifting quality left improved collaboration with partners such as product and design teams measuring quality, uptime, and cycle time etc. writing efficient tests improving build speed reducing handoffs shifting quality left improved collaboration with partners such as product and design teams measuring quality, uptime, and cycle time etc. Originally published at https://medium.com/vertical-slice/getting-software-engineers-to-do-the-right-thing-d42a0b923408 https://medium.com/vertical-slice/getting-software-engineers-to-do-the-right-thing-d42a0b923408

Stories, News & Mentions about SQWAD

Sqwad’s Stories on HackerNoon

Sqwad's News & Mentions