Chong Yu

Today I'm sharing this 🔥 video from our recent Engineering onsite at Demandbase HQ in SF! https://lnkd.in/eFAFR6QG We were so fortunate to host Zhengxuan Wu and Aryaman Arora, two upcoming AI researchers from Stanford University Department of Computer Science. Their approach to fine tuning, called ReFT, targets the representations emitted by neural net layers rather than the weights of the layers themselves. This makes it very flexible when using the same base model for swapping multiple fine tunes. The memory overhead is also very low for their intervention weights. Another highlight is how powerful it can be. One POC they showed forced the LLM to speak in only emojis exhibiting ReFT's power to control output. Control is crucial when, for example, you want to prevent your bot from offering customers unauthorized refunds!! If you're already familiar with LoRA, it should be pretty simple to use with only a few extra hyperparameters -- details in the video. Here's a link to the repo. Give them a star! https://lnkd.in/eJScNbzK Thanks again for letting us host your first external talk Aryaman and Zen. It's very exciting tech!!

17

4 Comments

Excited to share that today we launched Amazon Bedrock Studio in public preview. Really proud of the team for for all the amazing work. 🎉 Bedrock Studio is a web-based platform designed to turbocharge generative AI development journey. Here's what you can expect: 🔹 Rapid Prototyping: A seamless development experience with access to top-performing models models and key Bedrock features like Knowledge Bases, Agents, and Guardrails. 🔹 Simplified and Secure Access: Utilize your company's single sign-on credentials to seamlessly access Bedrock Studio, eliminating the need to broader access to AWS management console. 🔹 Collaborative Environment: Ideate, experiment, and refine your generative AI applications effortlessly with your team. Builders can kickstart projects in an isolated environment, inviting team members to collaborate and refine generative AI applications. Demo Video: https://lnkd.in/geGqrEgV Excited to embark on this transformative journey with you all! 💡✨ #AmazonBedrockStudio #GenerativeAI #Innovation #AIDevelopment

99

2 Comments

Come hear about our open source observability journey with Rodrigue Koffi and Marc Chene!

22

I reached 17 years at Google two weeks ago. Google has changed quite a bit over the years, though I won't get into that now. Sometimes, though, it's not the big things you've landed, but the smaller things that unexpectedly turned into bigger things. I frequently reflect on: - Android lint testing for Google apps - Fake "Save" button in Google Docs - The algorithm for Google Calendar's event layout - A couple of random test frameworks I created along the way, e.g. one that repeatedly prevented new crashes in Android OS itself - Some novel algorithms for detecting production issues in Android - Helping land XSS-free autoescaping for one of our most common HTML templating languages Many of these were things I started on a whim because I thought they might be valuable. Some of them I landed myself, but some were just started and took on lives of their own. Of course, I tried a lot of other things that didn't work, but it's important to try and experiment even when not everything works. What small things bring _you_ disproportionate pride?

291

19 Comments

🔥 𝐈 𝐚𝐦 𝐬𝐞𝐞𝐤𝐢𝐧𝐠 𝐝𝐞𝐯 𝐫𝐞𝐥 𝐩𝐨𝐬𝐢𝐭𝐢𝐨𝐧𝐬: 𝐡𝐢𝐫𝐞 𝐦𝐞 𝐭𝐨 𝐡𝐞𝐥𝐩 𝐲𝐨𝐮𝐫 𝐝𝐞𝐯𝐞𝐥𝐨𝐩𝐞𝐫 𝐜𝐮𝐬𝐭𝐨𝐦𝐞𝐫𝐬 𝐬𝐮𝐜𝐜𝐞𝐞𝐝! My name is Jacky, and I am 𝐚𝐜𝐭𝐢𝐯𝐞𝐥𝐲 seeking developer relations roles. My last post garnered some awesome convos with folks, so I wanted to put the feelers out again. 📮 𝐋𝐞𝐭'𝐬 𝐜𝐡𝐚𝐭: https://lnkd.in/eBHf5_DT Most recently, I worked for Pinecone as a senior developer advocate under the incredible Bear Douglas and Grigoriy (Greg) Kogan where I learned from the best of the industry to build + run a world-class developer advocacy team. I built and launched the viral Pinecone Shop The Look (https://lnkd.in/dxs6zJ5t) sample app that even got Mark Zuckerberg himself responding to netting over 300,000 impressions! I am ready for hire - I have my resume, work samples, and professional references ready. My core tenants as a dev advocate is: 1. Build 2. Communicate 3. Achieve company goals 4. Mentor 👉 𝘐 𝘸𝘰𝘶𝘭𝘥 𝘥𝘦𝘦𝘱𝘭𝘺 𝘢𝘱𝘱𝘳𝘦𝘤𝘪𝘢𝘵𝘦 𝘪𝘧 𝘺𝘰𝘶 𝘤𝘰𝘶𝘭𝘥 𝘭𝘪𝘬𝘦, 𝘤𝘰𝘮𝘮𝘦𝘯𝘵, 𝘢𝘯𝘥 𝘴𝘩𝘢𝘳𝘦 𝘵𝘩𝘪𝘴 𝘱𝘰𝘴𝘵 𝘵𝘰 𝘩𝘦𝘭𝘱 𝘪𝘵 𝘳𝘦𝘢𝘤𝘩 𝘮𝘰𝘳𝘦 𝘢𝘶𝘥𝘪𝘦𝘯𝘤𝘦𝘴. 𝘛𝘩𝘦 𝘫𝘰𝘣 𝘮𝘢𝘳𝘬𝘦𝘵 𝘪𝘴 𝘪𝘯𝘤𝘳𝘦𝘥𝘪𝘣𝘭𝘺 𝘵𝘰𝘶𝘨𝘩 𝘳𝘪𝘨𝘩𝘵 𝘯𝘰𝘸, 𝘢𝘯𝘥 𝘢𝘯𝘺 𝘭𝘪𝘵𝘵𝘭𝘦 𝘣𝘪𝘵 𝘩𝘦𝘭𝘱𝘴 𝘢 𝘭𝘰𝘵, 𝘵𝘳𝘶𝘭𝘺 🙏

3

My Amazon.com credit card pays for both Amazon.com and my personal AWS setup. I had a fraud report on my old card, and got a new one. Amazon.com and Whole Foods updated seamlessly. AWS didn't know. I just got an email about the fact that my payment method didn't work. Is this a sign that AWS and Amazon.com really are separate companies? When I worked there, switching between them was equivalent to switching jobs entirely (FULL interview loop, none of your internal performance was taken into account). OR is this a sign that integrating your businesses are hard, and even the biggest and "best" cannot do it? Note: My personal AWS bill is negligible. I use it to glacier my Synology and host my very bad personal website. The cost is << $5/month.

1

1 Comment

Well, it's official... Friday was my last day at Amazon, after over 16 years at the company! It was my first job out of college... so I hope to grapple more with this transition in future posts 😅 My first thought though... is that in the long run, you have to love what you do. I know Bezos catches some flak for the "work-life harmony" concept, but I find it to be true: work can often be a grind, but hopefully it's life-giving and energizing. (And as a chess coach, this parallels what I tell my students: chess can't be all about winning. You'll probably lose half your games too, so you have to enjoy the journey of improvement.) Reflecting on my career path at Amazon... I was admittedly a bit meandering and unambitious (only 2 promotions in 16 years), but what I found I valued most was autonomy. Finding those situations where I had Earned Trust and Delivered Results enough that I had a longer leash to trust my instincts and propose my own projects. Some of the best times were when I was very close with my business teams, and we could brainstorm together on what kinds of tools and tech could solve key issues for them, in ways they hadn't even imagined. But lately, my passion has turned more and more towards Education. I've had some opportunities to coach and teach in Chess, Math, and Software... and it's so rewarding to see those light bulb moments, as students reach for their full potential! And I firmly believe that tech and AI could automate the coaching process, making it widely available to more and more learners of all walks of life. To that end, as I'm taking some time to rest and recharge... I want to go deeper with AI tools and really get a feel for what they can do, especially for coding and learning. What tools or content have been most helpful on your AI journey? Please share! 🙏 If you've made it this far, thank you for reading! It's new territory for me, cutting my purple tether from the Amazon mothership... so I'm excited to be on this journey with you all. Even post-Amazon, it's still Day 1!

404

26 Comments

From "tunnel rat" to senior engineer with FAANG on the resume - a TL;DR of my salary journey in tech: 2005-2015: Worked for the MTA in NYC as a Track Worker. I was literally swinging hammers on the rails, laying down tracks. 2015: Attended General Assembly, a 3-month bootcamp, and learned how to code. Quickly resonated with Front End development. Was the 1st in my class to get a job offer and by the summer I was working at Accenture. This represented an $11k raise from the MTA. 2016: Gave myself a $22k raise and accepted a job at L2 Inc. My first startup! I learned React here as well. Good times. 2017: Gave myself a $24k raise and accepted a job at Transfix. To this day, this was one my my favorite companies to work for. This was a hard company to leave, but I was early in my career making up for lost time in the tunnels. I also spit a verse over a dope beat as my farewell email and sent it to the whole company. 2018: Gave myself a $48k raise - welcome to Bloomberg! At this point I thought I plateaued. I was now making more than 2.5x what I was at Transit. Talk about a life change! God is good. I had no plans of leaving until drastic internal changes happened. 2019-2021: Took a 3k base pay cut but... joined Meta and give myself a significant overall raise due to those sweet, sweet RSUs. God, is this real?! Some of the best working experiences I've ever had. What a ride. 2021-2023: Base pay bump of 30k - now I'm at Coinbase. And still, those sweet, sweet RSUs. Liquid cash is always the best! This would be the company that made me realize I had the "golden handcuffs" on. It's SO hard to walk away from RSUs, but my family would ultimately help me realize there's more to life than money. 2023 - present: 30k base pay bump as I joined the good ship Docker, Inc. Pretty much at the top level (financially) of being a senior engineer but the golden handcuffs are finally off. Join me and 🌻 Anna Miller this Thursday as we continue our conversations about my career journey. The whole point is to give people insight and encouragement on theirs. 🫡

102

23 Comments

If you’re building platforms for developers you have to watch what engineers are doing with output from your UX. Almost always there’s a cut and paste operation that you can make easier for them.

26

4 Comments

After 14(!) years at Google, I retired last week. (I reserve the right to retroactively call it "a sabbatical" :) Google encourages engineers to move about within the company. I was lucky enough to take a bit of a tour:  I started in Ads (and helped a little with the transition for DoubleClick). Then I joined Workspace (GSuite at the time) and helped rebuild (and launch) Sheets. Then led the development of the Sheets Android app. (At the time, it was ambitious to have such a complex app run entirely on a phone.) When VR had its first "moment" I went to Daydream and helped build Blocks (now open source!) and led Poly. Unfortunately, Google quickly lost interest in VR and defunded everything shortly after launch. (Sad trombone.) So I went to Area 120 and built Tables (now graduated as AppSheet Databases). My final stop was in Labs. There, I helped launch NotebookLM. And I led another not-yet-launched (LLM) product – that I *hope* will see the light of day (soon). I checked some stats on my way out. The engineering org -- the part that makes stuff, including PMs, UX, etc -- grew 10x in the years I was there. Half of the people who started before me were still there when I left (on my last day, I had been at Google longer than 95% of others in the org). When a company grows 1000%, things change. Every product I launched was harder than the previous one because of -- among other things -- increasing red tape. I understand the need to be more conservative as a company grows, but all the process does mean: more paperwork; less building. (Which is less fun.) Having said that, I still think Google is a great company to work for. Some of the best engineers, PMs, UXers and managers I ever met worked (and still work) there. I’m so grateful for the opportunity to have spent 14 years at that amazing place partnering with so many wonderful people.

1,242

81 Comments

I’ve been meaning to restart my blog, and I’ve also been wanting to dive into more distributed systems papers. So here’s me doing both: I just published a post on Amazon’s Dynamo, the distributed key-value store that laid the groundwork for DynamoDB. It’s a fascinating example of how distributed systems can be tailored for specific use cases. The blog is essentially my notes from reading the white paper. Any feedback would be greatly appreciated! #DistributedSystems #Dynamo #NoSQL #AWS #TechBlog #CloudComputing

21

🤔 Question for Software Engineers and Recruiters: In a world where the job application process has become a parody of itself, and carpet bombing open positions with resumes is easier than ever, what are some realistic ways to actually stand out and get noticed? Does it all boil down to who can best game the AI filters? Should we rely more heavily on our connections, or perhaps seek to build relationships with more recruiters in hopes that they'll direct attention our way? Short of becoming a LinkedIn influencer, there must be some practical steps to cut through at least some of the noise. What strategies have you seen work in the wild? *** shameless self promotion *** I'm looking for my next contract as my current one will be coming to a natural end soon. #JobSearch #Recruiting #SoftwareEngineering #CareerAdvice #LinkedIn

8

7 Comments

My thoughts on bypassing dumb biases when trying to get your next tech position... The "overqualified" bias: If you are having trouble finding your next job because folks keep saying that you are overqualified, trim some stuff off of your resume. Seriously, take 5 years or 10 years off. You do not, in fact, have to give your whole work history. If you keep getting the message over and over that you are over-educated, trim back on that. Your next tech position does NOT need to know that you have a doctorate in music or french literature. If it isn't helping - strike it. The "they'll say anything to get a new job because they were laid off and aren't working" bias: I had this happen to me one of the first times I was laid off. I had someone (chiropractor) very interested in hiring me. The interview had gone very well and on the way out the door he asked what my schedule was with my current employer because he wanted to schedule a follow up (I was doing contract massage therapy at the time). So I told him that my current gig had had to cut back on the therapists they had due to less clients scheduling sessions. It had been maybe 3 weeks since I had had a session and I had been cut because I was one of the newer folks. "OH, I see." That was literally the end of that. The interest fizzled up and died, right there. So... do contract work. The first time I was laid off from a tech job I had a friend that needed a web app set up for their business to help advertise a salon but couldn't afford regular rates. So we traded - I gave them a discounted rate and traded for some of the businesses services... yep got my hair done several times. I later found a similar contract with a friend trying to expand their existing website offerings and get better results in google. This was without the trade but it was helping a smaller business at a reduced rate. The point being that as I started getting serious interviews some months later, I had no gaps in my employment. No, I hadn't made good pay at either of those "jobs" but I filled in the resume gaps while my "day job" was mostly trying to get new employment. Not enough Experience bias: A similar thing can be done for folks who don't have enough experience. If you are working in something else but trying to get a job in technology/IT, moonlight or trade with a friend or local business at a reduced rate. You'd be surprised at how much value you can get out of a trade. The only caveat being, make sure you do your best to honor the agreement. Don't treat it as a throw-away contract -which is also why you can/should be picky when offering services at a reduced rate. You will very likely want this person to be a reference for you when you go to get your next job. Make that clear and only trade with someone that you believe will keep up their end of the bargain. If you have family with their own business - seriously... go do website work or IT support for them and list that. Good luck out there.

4

I've been hard at work, getting ready to stand up Pixel Valley! Got a lot done on the conversation engine and created a quick StreamLit app to test it. Doesn't look like much, but I feel this is a big deal because it is "not just another chatbot". As you can see here, I am having a coherent conversation with an entire group of agents, that are each responding or not responding as appropriate. This was created with off-the-shelf gpt-3.5-turbo. Instead of using expensive fine tuning, it is accomplished using a combination of clever prompt engineering, RAG, and AI function calling. It's getting close!

11