Gautam Shende

🚀 Making LLMs work for you - A Guide to Become 10x Programmer/MLE/DS I am sure many of you use LLMs for your day-to-day work. Mostly, I have seen people use LLMs while writing code (Copilot), or as an alternative to Copilot to catch bugs, but they have become much more useful than that, provided you know how to use them well. Here are a few ways I have been using LLMs in my daily work that are helping me increase my productivity: 📚 1. Quick Paper Assessment with ELI5 Summaries - Paper abstracts are called abstracts for a reason. Asking an LLM for an ELI5 summary makes it much easier to know if I want to read a paper. 💻 2. Implementing Research Papers in Code - Try giving an LLM like Claude/ChatGPT a research paper and ask to code it. You would be surprised. Tip: Claude performs much better for this task. 🔄 3. Upgrading Code Structure and Readability - Do you want to add typing hints when writing your Python code, or do you need comments in your code, or do you have a big function that you want to be converted to classes in your code? LLMs can do it all, and you would be amazed at the code-conversion power they have. ⚡ 4. Optimizing Code Performance - The thing with LLMs is that if you tell them what to do, they can do it well. For example, if you give them a non-optimized piece of code that runs correctly, they will be able to write it in a much-optimized way. For example, I just used an LLM to convert a matrix multiplication in PyTorch to sparse MM as I was getting an OOM memory error because one of the matrices became too large. I am not even joking, but that would have taken at least 1-2 hours for me to come up with. 📝 5. Automated Documentation - Got bored of writing documentation for what you did but have the code? LLMs also work the other way around. Want to write a summary of what you did in a week, you can use an LLM for that. Saves so much time that gets wasted in formatting. 🧪 6. Streamlined Test Framework Creation - Let LLMs handle the boilerplate for Unit tests & Integration tests while you focus on core logic. 🌐 7. Language-Agnostic Development - You can paste a function in Python and ask for its equivalent in Rust, complete with language-specific best practices and error handling. This dramatically reduces the learning curve when working with new languages and helps maintain productivity during language transitions. 🤝 Apart from all of the above, LLMs work well as mentors as well as interns. LLMs are great when you just need someone to bounce ideas with. I sometimes just talk to an LLM as I would to a colleague talking about the problems I have and how I can solve them. And they come up with the right ideas or send me in the right direction a lot of times. They also act as a personal assistant for free, who can deal with writing messages, creating charts, and doing all sorts of uninteresting work for you. 💭 What's your favorite way to use LLMs? Let me know in the comments! #MachineLearning #AI #Programming

46

Recently one of the stories regarding investing in Intel Corporation is going viral. So apparently a college math major got a $800k inheritance from his grandmom and he used $700k of that to invest in INTC hours before the Q2 2024 earnings were out. I know, a terrible choice in hindsight! 🙃 And he lost $200k of that within hours after the terrible report came out and management expressed cluelessness. That too on stocks, he was not even doing derivative leveraged play like options. 😅 This story is a very good lesson that stocks inherently are a risky investment, and it is good to be diversified to some extent. Some lessons to keep in mind: 1. Do not put all your eggs in one basket unless you know the business inside out. 2. Always do your due diligence on the company and the management. Company being well known is not enough, it is also about the ability of the management to execute. 3. Just buying low P/E stocks is not value investing and is not safe by any means. #investing #equity #stocks

26

4 Comments

Over the last 1 year, I've dived deep into the codebases of major projects like TensorFlow, MongoDB, Chromium, OpenSSL, LLVM, GCC, Kafka, Qt, CMake, and currently the Linux kernel. Transitioning from data structures and Development using Javascript based Tech Stacks to exploring open source C++ projects felt like stepping into an entirely new world. What I discovered were common practices and patterns that define production-level C++ code across these projects. Curious to know what they are? I've compiled my insights into a PDF—perfect for anyone looking to explore the Open Source Projects and contribute to the same. #C++ #OpenSource #SoftwareDevelopment #Programming

86

8 Comments

Intuit's recent decision to lay off developers, citing the need for more advanced AI skills, comes off as a disingenuous manoeuvre that sidesteps the truth. It's hard to buy into the narrative that these layoffs are truly about acquiring employees with better AI proficiency. In reality, it seems like a convenient cover for a cost-cutting strategy that involves offshoring jobs to cheaper markets. Think about it: the idea that current developers lack sufficient experience with AI tools like ChatGPT, Copilot, or other generative technologies is misleading. These tools have been part of the industry landscape for the past few years, and developers everywhere have been getting up to speed with them. The notion that there's a significant skill gap among developers when it comes to these tools is simply not accurate. The tools themselves, while helpful in certain contexts, are far from the magic solutions they are often hyped up to be. At best, they serve as fancy autocompletes—useful for streamlining some tasks but incapable of replacing the nuanced, expert work required for sophisticated software development. The real issue here is that Intuit's move undermines the trust and morale of its employees. Firing seasoned developers under the pretext of needing more AI-savvy replacements is not just misleading, it's a slap in the face to those who have dedicated their skills and efforts to the company. It creates an atmosphere of insecurity and distrust, where employees are left wondering if their hard work and loyalty mean anything in the face of corporate cost-cutting. Moreover, this strategy reflects a short-sighted understanding of what AI can achieve. While AI tools can certainly assist with certain aspects of development, they are not yet sophisticated enough to replace human expertise in creating complex, production-ready software. Believing otherwise is a risky gamble that could compromise the quality and innovation of Intuit's products in the long run. If the true intention behind these layoffs is to offshore jobs to cheaper markets, then Intuit should be honest about it. Hiding behind the AI narrative is dishonest and disrespectful to the employees who are being let go. Offshoring comes with its challenges, such as communication barriers and potential declines in work quality, and it is not the silver bullet it might appear to be. In an era where ethical business practices and corporate social responsibility are increasingly valued, such a strategy can erode customer trust and loyalty. #intuit #layoff #jobs https://lnkd.in/dv3qD9kS

4

1 Comment

Best time to buy Nvidia Stock? . . This is something really commendable from Nvidia!💯 This is a major milestone for the chipmaker, driven by the surging demand for AI technology and Nvidia's position at the forefront. The future looks bright for the world of AI, and I'm excited to see how Nvidia continues to innovate. Are you aiming to buy the stock after the 10 for 1 split seeing the future? Please drop some tips for all of us if you know about the US stock market. TIA! ☺️🤟🏻 #NvidiaNews #AITechnology #MarketCapitalization #StockSplit #TechIndustry

9

Distinguished Engineers (E9) at Meta make $4m/year. That's 16x the Bay Area median ($255k) Let's break it down [output = input x leverage] We all have 24 hours in a day, so there is a cap on [input] But hey, let's say they manage to have 2x-3x more focused time, so their [input] is 2x-3x There is still 5.3x-8x to go And that's [leverage] Here is the [leverage] they have: 1️⃣ Writing Every piece of text they write is clear to everyone, and they write a lot, for a lot of people 2️⃣ They create something from nothing And that something has clear value to the business 3️⃣ They influence other Engineers around them They are often described as having a "Midas Touch" in terms of their ability to uplift and develop other Engineers around them The best indicator is that the Engineers in their teams get promoted faster 4️⃣ They are well known and have an impact beyond Meta, promoting the brand 5️⃣ They have industry-leading experience in some areas There are rumoured to be <50 E9s at Meta

13

2 Comments

Gravitas | Google's Gemini AI fiasco triggers calls to sack Sundar Pichai | WION Google CEO Sundar Pichai has called the controversy around Gemini AI's inaccurate and biased responses as "completely unacceptable". In an internal memo to his employees, Pichai assured that his team was working on fixing the issues but investors are already calling for the CEO's resignation as a result of the controversy. The fiasco has sparked a bigger debate about AI. Can the public trust AI models? On this episode of Gravitas, we try to decode. #google #gemini #gravitas https://lnkd.in/guikXN4S

2

A recent breakthrough titled "Matrix Multiplication-Free LLMs" demonstrates a huge advancement in the area of Large Language Models (LLMs) by reducing computational costs. The authors have eliminated MatMul operations from LLMs, claiming to 10 times reduction in memory usage and a 25.6% increase in training speed, all while maintaining strong performance at billion-parameter scales. Paper link: https://lnkd.in/ggph8qXc #AI #machinelearning #deeplearning #LLMs

16

Google’s DeepMind has released a new method called JEST, which claims to train AI models 13 times faster and 10 times more power-efficiently than current methods. An AI model is only as good as the data it is trained on, and the time taken to train an AI model is proportional to the quantity of data. Naturally, since the new method JEST is 13 times faster in training, it has been trained on less data — which is of higher quality. The paper released by Google discusses optimizing the data selected for training the model. I found the solution to be quite smart: identify dependencies between data points and accelerate the training process by focusing on data points that match or have dependencies. Quality >> Quantity. AI systems demand a lot of energy and water for cooling today (annually ~ 4.3 GW, 0.5 litre water per 5-50 ChatGPT prompts) — JEST is a promising and positive sign for the AI world we are heading towards. #ai #deepmind #google

65

2 Comments

Last week at HackMIT, my team (Sanskriti Akhoury, Krish Modi, and Aly Shariff) and I built Runk: an AI-powered e-commerce platform that understands your preferences to display relevant reviews and products. How do we know your taste? Using autoencoders, we assign each user a profile embedding, which our own model updates as you leave more reviews. With these profile embeddings, we rearrange e-commerce search results to display products with the most positive reviews left by people most similar to you first. We also built a Google Chrome extension to interface Runk onto pre-existing sites like Amazon to re-rank these reviews based profile similarity. A huge thank you to HackMIT for making this hackathon an amazing experience. I also enjoyed meeting new people at Massachusetts Institute of Technology, Harvard University and Northeastern University! 👾https://lnkd.in/eyhU3nvK Outside of this hackathon, here’s what I’ve been up to this summer: • Built Guardian at HackThe6ix - an AI wearable for SOS situations. With a Raspberry Pi, Gemini Flash, and Hume AI, it contextually analyzes audio and video of emergencies and sends a summary to first responders. Collaborators: Sanskriti Akhoury, Daniel Hong, and Brian Zhang. 👾https://lnkd.in/ei5-Gu46 • Built Siren - an AI-powered dashcam. It analyzes red light camera runs, illegal turns, speeding, and other driving behaviors using YOLOv8 and algorithms built with reverse geocoding. 👾https://lnkd.in/eKMs2x5t • Joined Mimrr as a Machine Learning Engineer to work on their LLMs for automating code documentation. • Joined WATonomous as an Autonomous Software Engineer to work on self-driving cars. You can view these projects on my new website! 👾https://lnkd.in/e-hWsAxB

987

20 Comments

Some Googlers have published research on indefinite context for large language models: https://lnkd.in/gDHRc7pu I suspect this may be the mechanism for indefinite context underpinning Gemini 1.5.

3

RAG is one of the basic techniques to leverage large language models in combination with the custom knowledge base. In my recent blog post, I explore what exactly happens inside RAG-based solution, even before the first byte of the prompt is sent to LLM. #LLM #AI #ML #RAG #PromptEngineering

9

Good evening everyone, It was an opportunity to interact and learn from the two Product Operators at Rethink Systems headed by Shravan Tickoo. 1) Claire Vo, #Founder of ChatPRD, #CPO of LaunchDarkly, 2) Alisa Haman, #HeadOfGrowth, ChatPRD. Claire Vo has listed out certain attributes which are needed to become the great Product Operators. ✔ #RelentlessCuriosity - Curious to understand the things. ✔ - #WillingnesstoLearn - Learn from anyone whichever one do want to learn and embrace change. ✔ #BusinessAcumen - Build up your skills at your current position, keeping the Company's vision, understand the Economics, marketing and build the Organization Revenue and profits. ✔ #LowEgo - Builds a sense of good work culture and collaboration with cross-functional teams. ✔ As you grow up in the hierarchy of the Product Operations, you need to look at #Metrics that drives the business and achieving #OKRs. ✔ One of the top lines that Claire Vo has stated in discussion is "𝙋𝙧𝙤𝙙𝙪𝙘𝙩 𝙈𝙖𝙣𝙖𝙜𝙚𝙧 𝙞𝙨 𝙣𝙤𝙩 𝙝𝙞𝙧𝙚𝙙 𝙩𝙤 𝙨𝙩𝙚𝙥 𝙪𝙥 - 𝙤𝙣𝙚 𝙤𝙧 𝙩𝙬𝙤 𝙜𝙧𝙚𝙖𝙩 𝙥𝙧𝙤𝙙𝙪𝙘𝙩𝙨 - 𝙗𝙪𝙩 𝙩𝙤 𝙗𝙪𝙞𝙡𝙙 𝙖𝙣 𝙚𝙣𝙜𝙞𝙣𝙚 𝙩𝙝𝙖𝙩 𝙘𝙤𝙪𝙡𝙙 𝙙𝙧𝙞𝙫𝙚 𝙗𝙪𝙨𝙞𝙣𝙚𝙨𝙨 𝙤𝙫𝙚𝙧 𝙖𝙣𝙙 𝙤𝙫𝙚𝙧." Alisa Haman has expressed her thoughts ➡ Never say "That's not my job". Sometimes Product Manager has to go beyond ones duties and make people aligned even when its not your job. ➡ With the intervention of AI technology in the current market, the PM domain is changing, especially with ChatPRD. Thanks to Shravan Tickoo for always coordinating with the great Product leaders . 💐 Thank you Claire Vo and Alisa Haman for sharing your insights that helps to build products from 0 to 1. 💐 Thanks to all the folks - Nadisha Garg, VARUN KHANNA, Mittali R., Kevin Thomas, Nikhil Yadav and Garima Arora for great questions. #ProductDevelopment #artandproductmusings #ProductLeadership

27

9 Comments

Read some papers about how sparse feature/embeddings are being used and optimized in recent day's recommendation system, especially in the era of LLM. Summarized my learnings and thoughts in this blog: https://lnkd.in/guZb4hXU Thanks for the great discussion with Yunzhong He and Yitong Zhou

74

1 Comment

While Gen AI engineers in India are rightly concerned about #hallucination in LLMs, yet most would rarely appreciate that genius folks in our own country evolved extraordinary measures to arrest hallucination thousands of years ago, which included sophisticated #guardrails which were to be 'assimilated' in the brain itself, impacting the way it perceived reality and 'generated' content (Isn't the brain one of the most hallucinating natural 'LLMs' after all ?). This was not just to ensure that facts were rendered correctly. The attempt was to prevent disasters of another kind and of a much more serious nature. After all, language in India was considered an important vehicle for connecting to the divine, and not merely for communicating with one another. And we often fail to realize this #opensource framework, so to say, was built upon by different intellectuals in different parts of the country, and India ended up having several wonderful languages in different regions, languages which look different, but have a lot in common. Did you not realize that no matter where in India we are from, names of most of us can be traced to this one language? https://lnkd.in/g4Njx4mG #llms #nlp #generativeai

17

3 Comments

My take is that the training process performs a kind of negative sampling to filter out samples that have low information, hence improving training efficiency.

5

3 Comments

I don’t think people appreciate the immense implications of this work, especially in the context of applying AI across sensitive NatSec info. Imagine a future where GenAI applications are able to maintain 100% end-to-end encryption of the data it handles, down to the GPU/inference level. It could drastically reduce cybersecurity risks and therefore the barriers to getting these apps authorized. Really exciting stuff.

30

4 Comments