Wei Xiao

Did you know there are no "roles" (SDE, SDET, TPM, …) when you get to L8 IC (Senior Principal) at Amazon? It’s one of Amazon’s peculiarities. [Side note: in general, Amazon SDEs tend to be "jack of all trades" where we pragmatically do whatever is necessary to make our organization successful at that given moment (within reason). In fact, I believe being well-rounded and able and willing to wear a number of hats is one of the things that makes me valuable to an employer. We don’t over-index on "that’s not part of my role doc." Other companies where I worked (eg Google) have very clear swimming lanes and people stick to them pretty religiously, eg. "oh no I don’t talk to customers, I’m an SDE and talking to customers is for TPMs." Whereas at Amazon I never once had a manager tell me, "what you did (although impactful, important and necessary) was not really your job." ] My little side note about this critical piece of amazon culture explains why as you get to L8 the concept of a "role" no longer really makes sense. The boundaries between roles get fuzzier and fuzzier as you go from Senior Engineer (L6) to Principal Engineer (L7), and they become unimportant when you get to L8. We’re all just “Senior Principals” - whether you came via the SDE path, the TPM path, or in my snowflake case, the SDET path. In fact I think most L8s are snowflakes in that their path to L8 was unique and the value they bring to the company is also unique. In phonetool (our amazon directory) you can have whatever unofficial title you want (also within reason...), so I chose to keep the "SDET" title to reflect my background and passion for testing, engineering and operational excellence, and developer tools. I also hope that by keeping the title, others in my (old?) role can see that SDETs at Amazon do not have a "ceiling" in their career. And lastly, it's a fun conversational piece ("you're an L8 SDET???") (The image here is "Peccy" - a beloved little icon that has represented Amazon's Peculiar Ways for years) [Edit: Strictly speaking there are a few L8 IC roles that do have a different job code, like PMT or Scientist - but I think we generally view all L8 ICs in the same lens]

929

41 Comments

Large Language Models (#LLMs) are evolving rapidly, and the demand for running them at scale is skyrocketing. #Ollama is stepping up to meet this need with its official container image on Docker Hub, excelling in local #deployments. However, there’s a catch for containerized environments: downloading #models requires the ollama serve command to be running within the #container. This extra step complicates automated deployments via #CI #CD #pipelines, making seamless model access challenging. To take your #model to production, it needs to be #container-ready—a feature Ollama has yet to fully support. The gap in #AIOps for smooth deployments gets blurred with the inference services available as an easy way to take it to production. Goal is to overcome obstacles and deliver solutions 😉 so here is to get you started on having your (O)#llama #production ready in a private #deployment from your #local to #cloud all in #oneshot . Enjoy Medium: https://lnkd.in/g9jqBQ8d dev.to: https://lnkd.in/gi_gGwam #AI #LLM #Ollama #llama3 #phi3 #mistral #S6overlay #AIOps #CI #CD #oneshot #container #docker #podmanEnjoy

11

/*********Mentorship**************/ I was watching today Andy Jassy's video (https://lnkd.in/gRV2mGPr) on career advice. He shared one of the very important thing on Mentor's. Mentor's overall objective is too help you grow. They let you do things which you even think you can not do and then give support if needed. They give honest feedback on areas of growth.

1

It is more efficient to bring logic to the data than to ship data to the logic. This is especially true for "real time" RAG scenarios where OLTP data needs to be searched and processed in real time. Azure Postgres now has the ability to generate vector embeddings in real-time within the database. Check it out! #postgres #azure #ai #copilot #rag #llm #postgresql

25

LLMs and LangChain: The Dynamic Duo 🦸‍♀️🤖 Just finished a fun project where we built a Q&A bot using these two powerhouses. It's like having a personal AI assistant that's always ready to answer your burning questions! 🔥 Here's what I learned: LLMs are wizards: They can conjure up answers from thin air (or, uh, the internet). 🧙‍♂️ LangChain is the glue: It holds everything together and makes it easy to use. 🔗 Endless possibilities: This tech can be used for anything from customer support to writing catchy social media posts. 📝 Want to see it live (obvio a test server link 😛 ): https://lnkd.in/dZNAVCSV Note: it's a render link, they will take down the instance because of no use, you can hit the link and wait for two minutes, they will spin up the instance and you can try it out. Want to see the code? Check it out here: https://lnkd.in/dzjh4r47 #AI #MachineLearning #NLP #LangChain #LLM #Python #Developer #FunProject"

22

Thats the exact explanation I gave to a group of investors last week. (… it’s just Precision/Recall, attributes vectorization as “embedding”, K-Nearest Neighbor, User Search Intent, DAG, ML algorithms, etc all over again …). LOL and they asked what’s new this time around. And I was like not much, ML is still the backbone, and AI is the inference and actions based on the model. It works better when there is a feedback loop in a closed system. But the current state of the art like ChatGPT is based on scraping the Internet (garbage in, garbage out). Kinda like early search engines. What is a better analogy is that NVIDIA/Snowflake etc are selling tools, aka pickaxes to prospectors. The gold is up to you to find. And the opportunity right now is to monetize industry specific (healthcare, travel, financial services) clean data to build models with (aka a treasure map for the prospectors). From an enterprise perspective, what’s out there today aka generative AI isn’t very useful. Too much hype. To build useful engines will require an effort similar to search revolution 2 decades ago. From

16

3 Comments

Next week, I will be the Principals of Amazon (PoA) talk presenter. It would be fun to have folks join (and ask good questions!) if you are an Amazonian.   Talk Abstract: Code reviews can be a painful ordeal, leaving both reviewers and authors feeling frustrated. But it doesn’t have to be that way! In this practical talk, we’ll explore techniques to transform code reviews into an enjoyable learning experience. You’ll learn how to give and receive feedback constructively, focusing on the code rather than personal attacks. You’ll gain strategies for structuring feedback, addressing comments, and handling disagreements without devolving into office brawls. Whether you’re a junior or mid-career developer, this talk will equip you with the skills to make code reviews a collaborative, educational, and (gasp!) enjoyable process. Join us to get a stress-free ‘ship it!’ on your next CR.   Here is a sneak peek at what we will cover. You might notice a few Principal Tenets in here. I use these daily in my work, in addition to the more famous Amazon Leadership Principles. The Principal Tenets are a framework for thinking about my work as a Principal Engineer and always raising my game. You can learn more about them here: https://lnkd.in/esPSssWv   Code Reviews that Don’t Suck   Learn, Educate, and Advocate You are NOT your Code Start with a Self-Review Consider Team Standards Check the Obvious Things Make it Atomic / One Change per Commit Respect what Came Before Make it Readable Provide Clear Context Go Talk to a Person, in Person Work Hard; Have Fun; Make History   How about giving Code Review feedback that doesn’t suck? Lead with Empathy Don’t be too Harsh But Be Clear Use Radical Candor Give a Full Review Be Prompt… or Delegate Act as a Co-Owner Coaching a New Engineer Hold High and Consistent Standards Use Comment Conventions Include Praise

300

9 Comments

Exciting news! Our team has released a new feature that I'm thrilled to share with you all. Auto-Association is now available to help our customers connect their data to our AWS Supply Chain Data Lake. Thanks to Generative AI, powered by Amazon Bedrock, we can read a variety of source data and automatically build the SQL transformation recipes needed to convert this data into our canonical data model. I'm particularly proud of this feature and want to extend a special thanks to my manager, Jyothi Bodas, for leading the project. If you're interested in learning more about Auto-Association, check out the demo video Jyothi created. I want to thank the entire team for producing such an awesome feature for our customers! Alok Mehta, Asim Joshi, Christie Molloy, Colin Ashfield, Leonardo Murri, Michael Walsh, Parshant Chowdhary, Shalini Thinderu,Tom Nolan, Vita Grebeniuk and the host of other team members that made this happen. I also want to send out a thanks to our leaders- Ameet Vaswani, Diego Pantoja-Navajas, Dilip Kumar, Alex Dunlap, and to our Customer Success team led by Tariq Choudry. #aws #amazon #amazonbedrock #AWSSupplyChain #day1

104

8 Comments

We're in the pets phase of LLMs right now, but the cattle phase is coming and it'll change everything. Yes, again. What do you mean "pets" and "cattle"? Servers used to be our pets. We bought, named, cared for, and persisted them individually. If we wanted a new one, it was a big commitment. If one went down, it was a big loss. But then servers became cattle. Today we manage them as a group of commodity assets. We want one, we clone it. If one goes down, we replace it with another. This change delivered an era of distributed, high availability applications with limited downtime. Cheap storage, containers, cloud, clusters all met riding on top of Moore's law. And machine learning changed too. Computer vision was first. We went from small, neural networks running on CPUs via Matlab to a larger networks running in on GPUs via C++. AlexNet kicked off the deep learning revolution, eventually Transformers in NLP, and then ChatGPT. So, today we call our LLMs by name but this is just the start. There will be more LLMs variations than there are people on Earth.

25

1 Comment

Project update! My application tracking extension is now doing fuzzier matching and injecting new information into LinkedIn (and other sites) to keep me from doubling up on applications. Next up: local storage and bulk import/export operations...

12

2 Comments

Anthropic's latest Claude AI model is taking human-AI interaction to a whole new level! 🖥️🤖 The new Claude 3.5 Sonnet can now use computers just like a human would - by looking at the screen, moving the cursor, clicking buttons, and typing text. This is a major leap forward in AI capabilities. Key features: Operates based on screenshots, emulating human computer use Can perform complex multi-step tasks across different applications Self-corrects and retries when encountering obstacles2 Accessible via API for developers to build with Why it's game-changing: This opens up incredible possibilities for AI assistance in data entry, form filling, software testing, and automating repetitive computer tasks13. Imagine having an AI assistant that can seamlessly work alongside you, handling tedious computer work while you focus on higher-level tasks. While still in beta, this technology demonstrates how AI is rapidly advancing toward more human-like interaction with digital interfaces. As it develops further, it could revolutionize how we work with computers and boost productivity across industries. The future of human-AI collaboration just got a lot more exciting! What do you think about this new capability? #AIInnovation #FutureOfWork Here is a video about it: https://lnkd.in/gh8aNYEu

6

Over the course of each month, I spend a few moments each day or so collecting interesting links, and then, at the end of the month, I bundle them (with the help of some nifty automation scripts) into an easily digestible reading list. Here's everything that I found interesting in May 2024 about the irresponsible use of machine intelligence: #qfm #quantumfaxmachine #blog #irresponsible #ai #genai https://lnkd.in/e2uE49up

10

Amazon-to-Google-to-Amazon #a2g2a thought of the day: the way Amazon thinks about its Principal Engineers is unique - and I love it. “I always thought of you (and my other Principal Engineers) as Amazon employees that were on loan to me.” Yesterday I was catching up with my good friend and ex-manager Dan Sommerfield. Dan promoted me to Principal Engineer back in 2014 and was my manager for a number of years. As we were laughing about some of our shared past, Dan said that and it stuck with me. A few weeks ago I posted in my a2g2a series about Swimming Lanes. I’ve also posted in the past about the Amazon Principal Engineering community (see links to both of those in my comments). Today’s reflections combine those. Amazon Principal Engineers tend to have a healthy disregard for organizational and role boundaries. And Amazon’s DNA gives them the freedom to do so. At these big faang companies, for better or for worse, most orgs tend to optimize for themselves. It makes sense in terms of business needs: they need to move fast and expand their business. But this is often how we end up with duplicative work siloed usually at Director level. I’ve seen, time and time again: Principal Engineers look across organizations and connect dots that others don’t. The Amazon Principal Engineering community is a giant organic mesh. It worked incredibly well as a big mesh when I became a PE in 2014 and we were about 250 PEs in the entire company, but I’m continuously amazed that it still works in 2024 at today’s scale. This is a testament to the culture that some of our founding figures set decades ago- and our Distinguished Engineers continue to carry the torch today. It’s not that this doesn’t happen in other companies; it did happen at Google and Microsoft too. But it’s so intrinsically embedded in the DNA of an Amazon Principal Engineer that it’s a lot more natural and expected here than anywhere else where I’ve worked or considered working.

592

24 Comments

"Your main job is to make your product win." I think this should apply to any product that your company makes. I also like what he says following this statement about what is NOT your main job. I will summarize here: "Your main job is not to make your manager happy, it is not to make execs happy." Also, note, this main job doesn't only apply to PMs, but also to senior Engineering Managers, Tech Leads, and Designers. #makeyourproductwin #leadership

5

2 Comments

🔍 Demystifying Distributed Systems: Raft's Secret Sauce for Reliability As an engineering manager, I've seen firsthand how critical consistency is in distributed systems. Today, let's unpack the Raft consensus algorithm's approach to leader election and consistency. Trust me, it's simpler (and cooler) than you might think! 🎭 The Players: Leaders and Followers Imagine a high-stakes game of "follow the leader" where servers are the players. In Raft: One server is the leader The rest are followers But what happens when the leader stumbles? 🏁 Leader Election: A Race Against Time The leader sends regular heartbeats to all followers If followers don't hear a heartbeat, they call an election Followers nominate themselves and request votes The first to get a majority vote wins and becomes the new leader But here's the kicker: Not just any follower can become leader. They need to prove they're up-to-date enough to lead. It's like ensuring the new team lead knows about all the recent project decisions! 🔒 Consistency: Keeping Everyone on the Same Page Once we have a leader, how do we ensure all servers agree on the data? All changes go through the leader The leader replicates changes to followers Once a majority confirm, the change is committed The leader then tells everyone it's official Think of it as a team decision: once most agree, it's locked in, and everyone gets updated. 💡 Why This Matters: In the world of microservices and distributed databases, Raft's approach ensures: There's always a clear decision-maker Data remains consistent even if some servers fail The system can recover quickly from failures Real-world impact? Imagine an e-commerce platform that never loses your cart items, or a banking system that always knows your correct balance, even during server updates! 💾 Real-World Database Examples Different distributed databases handle consistency in various ways: 1. etcd & Consul: Use Raft for strong consistency. They're like a team that always makes decisions together. 2. Apache Cassandra: Uses a different approach called "eventual consistency." It's more like team members working independently and syncing up later. Pro: Super fast writes and high availability Con: Might temporarily have inconsistent data across nodes 3. MongoDB: Offers a spectrum of consistency options: In replica sets, it uses a consensus algorithm similar to Raft Also provides eventual consistency options for faster performance 4. CockroachDB: Uses Raft for strong consistency while scaling horizontally. 🤔 Question for you: Have you encountered consistency issues in your distributed systems? How did you tackle them? #DistributedSystems #Engineering #TechLeadership #Raft 🔗 Part 2: https://lnkd.in/dRhU3B63 Join the Journey with Vikas K. 🔥 Follow for a World of Insightful Content!

9

Two facts: First: Jack and the Storefront team handle some of the most interesting problems at Recharge if you're into caching, synchronizing, and lots of data puzzles to solve. Second: I work with the team as Principal so we'd get to have some fun while we do the first part.

17

I get it: y’all have AI fatigue. And yet I’m going to quip on my personal experience, this time with C#. Admittedly, I’ve never coded in C# (vs Java), nor I am familiar with specific language constructs. No biggie! Comes Amazon Q + VSCode. 1. AQ was able to explain code pitfalls. 2. AQ was able to point out memory leaks. 3. AQ recommended (defensive) code changes. I’ve tried it with defects and stories type of work (new work vs rework). With no prior experience in programming language, with a bit of formal CS education, and with a bit of brain and logic, and a good peer review!, was able to complete the work in hours vs days. If you are in the fence or you are already an expert - I get it, you may find it boring and amusing. But don’t pass on it, give it a try, confirm with your own experience what works for you. I guarantee that Amazon Q will cut mundane code boiler plate generation time and basic problem finding time by 50% at least. https://lnkd.in/gV54G3xs

4

Curious and slightly unsettling experience with chatGPT during a novel review processl (it has been a long time since the publication of my last fiction book!). I uploaded 4 chapters using the premium version and requested summaries of the main characters' names, surnames, and descriptions. One character intentionally had no surname, but chatGPT provided one. After pointing out the error, chatGPT apologized, acknowledged the mistake, and offered a different surname. When I inquired about the repeated incorrect surnames, chatGPT explained its reasoning: Certainly! The reason for the earlier errors with the characters' surnames is a mix of factors: Initial confusion in searching the document: When I tried to retrieve the names from the original document, the search didn’t extract the correct information, leading me to create plausible but incorrect surnames instead of relying on the actual text. Autogenerated names: In cases where I couldn’t locate the correct surname, I defaulted to generating surnames that seemed fitting for the context or the character’s first name. This was a mistake, as I should have used the accurate details from the text. Imprecise searches: My early attempts at searching the document didn’t yield the right results, so I made assumptions instead of double-checking the content properly. #chatGPT #AI

21

2 Comments

I just realized: we should start saying "build memoization" rather than "build caching." We aren't really talking about least-recently-used or time-to-live algorithms, we're talking about "don't bother recompiling this piece if none of its inputs changed." (It has been puzzling me that some people are so deeply fearful of "build cache" as a concept, and honestly I think it's mostly because most build systems have non-hermetic state scattered all over the place. Maybe a slightly more precise term for the type of caching we're concerned with would clarify the semantics in question.)

83

9 Comments