Nishita Agarwal

Our Kumo.AI blog post on Hybrid GNNs is out. https://lnkd.in/gpzr8TiF We have seen success with our Hybrid GNNs across several marketplace problems. Hybrid GNNs are a novel innovation at Kumo.AI because they model both repeated and explorative user interactions within a single GNN framework. The GNN automatically infers whether a user tends to do repeat purchases or tends to be more exploratory (ie buying new unseen items) and learns a repetition scalar per user. What we love about this approach is that it takes away from the standard recommendation systems architecture where you have multiple candidate generators - some for new unseen items, some for repeat behavior and then have to combine them in a final multi-objective optimization framework. Such systems which are a combination of multiple systems are hard to maintain, re-train and operationalize. On the Kaggle H&M challenge where most winning teams usually have approaches that require large amounts of feature engineering and the final system is a complex ensemble of approaches needing months to build, the Kumo.AI approach which is a 3 line predictive query that runs a Hybrid GNN under the hood is top 1%. Thanks Matthias Fey and Weihua Hu for the algorithms and innovation.

158

2 Comments

Congrats the team on ICML acceptance of the Wukong paper: https://lnkd.in/g-MyahNH, which represents a breakthrough of recommendation model design and efficiency. Recommendation models traditionally suffer from not able to utilize the full computational power of GPU (only 5-10% of peak GPU FLOPS while LLMs can reach >50%), due to their sparsity. Wukong is first of its kind by pushing toward 40% GPU peak FLOPS though model-infra codesign. One of my teams consisting of GPU experts is building a high performance model component libary where we’ll provide highly optimized model components (eg sequence summarization, x-sequence interaction, x-GPU layout scheme) that could be composed - with the goal that models built from this library have high performance out of box to scale the model-infra codesign effort. We are also scaling training beyond the limits that are typical of recommendation models - from hundreds of GPUs to 10s of thousands. If this is the kind of work that excites you, come join us!

264

4 Comments

Reforge reached out to me a few months ago and asked if I would be willing to create a few artifacts and share them with their community. One of the artifacts I created was a PRD for a made up feature for Netflix: Calendar reminder for live stream PRD I have improved this PRD over the years, and while there is always room for improvement, I hope you find it useful for your PM work https://lnkd.in/eVWiGSXS #productmanagement #PRD #growth #reforge

25

https://lnkd.in/gxJWb4hA Just testing! :) I "wrote" this short article with help of AI - in minutes, but it is about an important difference, that I would like to highlight, between both awesome APAC hubs: HK and SG. For example it would be great if HK introduces LTVP visa for non-immediate family members and it would be also great if SG allows spouses to take up jobs again. The issue is complex, but worth a read!

6

Today, every founder wants their CTO to be hands-on Why? Companies today are strapped for cash It is the need of the hour to get the the most out of everyone including CTOs As a hands-on CTO myself for over a decade,  I am blessed to practice my craft everyday! (I highly recommend all CTOs and VPEs try "hands-on" engineering) The benefits are: → Rich interactions with the engineers, because I can speak their language and work alongside them → Easiest way to uncover inefficiencies in the workflows, and put plans in place to fix those → Fastest way to not only know the latest tech trends, but have real-time, hands on practice to understand the pros and cons → Help onboard new engineers and along the way improve that onboarding process itself to be highly efficient → Make sure some overzealous engineers don't go bike-shedding :) → Market to prospective hires with pride and confidence TBH I am hands-on for the sheer joy of writing & deploying code  and creating value out of thin air Why do you want a hands-on engineering leader at your growing startup?

61

3 Comments

🌟 I've dreamed about this day. After I switched to product management, one of the resources that Siddharth Gopi shared with me was product hunt's PRD. That's when I first discovered this insane platform where people could share the products they were working on, get early user feedback and find out if it was all worth it. I write this as I prep for my own product hunt launch, after almost 9 months of working on my own product. I'll share more about the journey soon, but for now, I'd love to have your support. My product, howsthisgoing.com, is launching to the world on December 16th. If you're curious to learn more, head over to the Product Hunt page and get notified: https://lnkd.in/gwQZ_yvG

73

17 Comments

🚀Hi Folks, I'm Taking on the Product #Teardown_Challenge along with Arindam Goswami! 🚀 As part of an exciting challenge by Product Space, I’m stepping into the role of Senior VP of Product at #Zepto Quick Commerce! Zepto’s already crushing it with 10-minute deliveries across 30 cities in India, but there’s so much more potential to unlock! 💡 Here’s the mission: ➡️ Increase Zepto’s Average Order Value (AOV) to ₹600 🛒 ➡️ Boost product discovery and engagement to encourage bigger, more frequent orders 🔍 ➡️ Elevate the purchase experience to keep customers coming back for more 🔄 And I need YOUR help to make it happen! 🙌 🎯 Your insights on quick commerce platforms like Zepto, Blinkit, and Swiggy Instamart are key to shaping these strategies. I’ve put together a short survey, and your feedback will be crucial in tackling this challenge! 💥 Thanks for your support—let’s crack this together! 🙌 Link to the form - https://lnkd.in/g5v2y4wR #ProductTeardown #QuickCommerce #ProductManagement #ProductSpace

9

2 Comments

Don't Steal the Exciting Work: A Drunk Lesson about Leadership I was in Hong Kong last time, having a beer with Franz, the CTO of Covertech. We were geeking out over tech stuff, getting progressively drunker, and somehow, amidst the buzz, I ended up getting a mini MBA from him. One thing Franz said stuck with me, even through the haze of alcohol: "If you pick up that awesome piece of work, it's a disservice to your team. You're stealing the opportunity." At the time, I didn't fully grasp what he meant. But hey, everything sounds profound when you're drunk, right? It wasn't until a few months later that I found myself in a situation where his words suddenly clicked. As a leader, it's your job to spot those sparkly opportunities and hand them over to your team. People are most motivated when they're working on something exciting and challenging. And trust me, as a former developer, I get it. I want to sink my teeth into those juicy projects too. But here's the thing: I'm a manager now. My role has changed. If I hog all the cool stuff for myself, I'm stealing the chance for my team to grow and shine. It's not about me anymore. It's about making sure the work aligns with the company's highest interests and giving my team the opportunity to knock it out of the park. So, here's the lesson: Don't be a selfish leader. Don't rob your team of the chance to do something amazing. Your job is to identify the right opportunities and trust your team to run with them. It's not always easy to step back, but it's necessary if you want to build a culture of ownership and empowerment. The next time you're faced with an exciting project, remember what Franz said over that beer in Hong Kong. Resist the urge to take it on yourself. Instead, give it to your team and watch them rise to the challenge. That's what true leadership is all about.

15

5 Comments

A Day at the Apache Kafka 🚀 Today, I had the pleasure of attending an in-person Apache Kafka hosted by CRED in Bangalore. Here’s my feedback on this incredible event: 📍 Venue: Cred Aero, Bengaluru 🔹 Agenda Highlights: - @Vikas Sharma , CRED : Talk: Running Kafka Optimally in Production: Practical Usage Feedback: Vikas shared practical insights on optimizing Kafka in production environments, with valuable strategies for resource management. - Udayaram Kammara , Bazaarvoice : Talk: How Bazaarvoice Utilizes Kafka for Both Stream and Batch Processing Scenarios Feedback: Udayaram's session on Kafka's use at Bazaarvoice was enlightening, covering real-world use cases and challenges. - Diptiman Raichaudhuri , Confluent : Talk:Stream Processing for the Non-JVM World: pyflink and kafka-python Feedback: Diptiman's explanation of Apache Flink's architecture and the demonstration of kafka-python and pyflink were fascinating and beneficial for non-JVM users. What I Learned: - Effective strategies for running Kafka in production. - Real-world use cases and challenges in stream and batch processing. - Leveraging pyflink and kafka-python for stream processing. Overall, this meetup was a treasure trove of knowledge and inspiration. The sessions were well-organized, and the speakers were exceptionally knowledgeable. I’m excited to implement these learnings in my projects and continue exploring the potential of Apache Kafka and related technologies. 🙌 #ApacheKafka #KafkaMeetup #CRED #Bangalore #EventStreaming #StreamProcessing #TechCommunity #Networking #Python #Flink #DataPlatforms

24

2 Comments

who are we building for on ton & why? - existing onchain user base - user base onboarded via recent airdrops - apps looking to monetise but lack licensing + infra - casual user base that wants to pvp on their internal chat without any global restrictions data to back my thesis: 👉 there are 22m activated wallets 👉 there are 600K+ daily active wallets gas is almost ~$0.025 the onboarding stage on ton will soon turn into the adoption stage, that's what are we building for.

22

5 Comments

Bytedance founder same spirit as Nvidia founder, don't you think so “[Zhang] Yiming can see that ByteDance needs a new growth engine after Douyin and TikTok,” said one person close to Zhang. “He is always thinking about what is coming in the next five years, what can extend the company’s future.” Feel free to comment ,like,forward and subscribe for more cybersecurity info/insight At qicyber.com Refer to news URL for more details #CyberSecurity #DataProtection #BusinessContinuity #formintiumit #formintiumtech #QiCyberCyberSecurity #5CyberCyberSecurity

1

I've been a tech enthusiast and programmer, in one way or another, since I was 6, nearly 42 years in all. Developing software, professionally is so much more than programming, and I've been at that for nearly 30 years. Along the way, I've worked in tiny-startups that scaled out into important players in their field, like Upvest, and Cmed Technology. I've worked at a few startups that didn't make it too, and a few multinationals (Thomson Reuters, HDI Group and Rentokil Initial) and core-tech firms like Canonical. One side effect of all that time is people start to ask you about what you've learned. Kovid Batra was kind enough to ask me about some less technical and more strategic aspects of that: product thinking and platform teams for the groCTO podcast. I hope there's something of value here, some things I learned the slow, hard way :-D

31

2 Comments

How you can make your LLM applications better!

5

Building products as an **#indiehacker** is challenging - not just technically, but emotionally too. I recently witnessed this firsthand through Jun's (https://x.com/shenjun1995) journey after he left the Amazon AWS team in Beijing. Jun's story is a testament to the resilience required in the indie maker world: 1. He let go of a side project he'd worked on for six months, following my advice to pivot. 2. Instead of rushing into new ideas, he embraced core #indiehacker practices:   - Validating before coding   - Marketing from day one The results? Jun made his first internet dollar in April 2024. His reaction? Tears of joy and a savvy marketing move: "I'm recording my emotional response for Twitter. It'll get traction!" That's when I knew: Jun had truly become a **#solopreneur**. A month later, Jun shared his rollercoaster ride: - Excitement led to creating 10+ articles about his product (https://lnkd.in/ewU56z3w). The harsh result? Zero conversions. - After a brief struggle, he pivoted again. With no prior experience, he created a promo video. The outcome? Traction and returning customers! - Next challenge: Reddit posts led to constant bans. Another setback to navigate. TL;DR: The #indiehacker journey is an emotional rollercoaster. Remember: - There's no single magic decision that solves everything. - Success isn't a jackpot; it's a series of correct choices accumulating to critical mass. - Manage your emotional energy to keep building those fractions of success. To all the #indiemakers out there: Keep pushing forward. Your "overnight success" might be just around the corner! #entrepreneurship #startuplife #buildinpublic

7

Project Astra: The Future of AI at Google Check out this awesome demo showcasing the capabilities of Project Astra, the next generation of AI from Google. https://lnkd.in/g_wiQ5cq

11

A few weeks ago, Hemant Mohapatra, partner at Lightspeed and I started discussing the evolution of agentic systems and realized there wasn’t a clear, canonical architecture to address the complexities of scaling these systems. So we set out to explore: 1/ What are the key infrastructure components required to enable agentic systems at scale? 2/ What role does UX design play in ensuring adoption and reliability? 3/ How can a multi-agent system handle real-world use cases like customer support? 4/ What are the opportunities for builders and challenges for enterprise adoption? How to think about differentiation? This article, "Endless Homers, Infinite Agents", is the first part of a series where we tackle these questions and share actionable insights. Feedback and comments welcome —hope you enjoy reading it as much as we enjoyed writing it! Link to the article: https://lnkd.in/g_q8XkcT

188

11 Comments

tl;dr - I am very excited about the future of AI video generation. Here is a mini music video I have created for my recently released song in < 1d with Runway Gen-3 Alpha: https://lnkd.in/gpmFBRBF == Background As some of you may know, I've poured my heart and soul into a music EP over the past two years (linktree: https://tr.ee/QWJiIOgRrz). One of the most thrilling aspects of this journey is bringing the music to life through video, enhancing the song's message and impact. Last year, I went to Istanbul and recorded a music video: https://lnkd.in/gyknijzU. I loved the outcome and am so happy about it. However, it required many hours of flight, a jetlagged version of me, a time crunch to finish shortly, and a huge barrier to entry. Now, you might ask how it is related to GenAI. Using Runway Gen-3 Alpha, I created a mini-video for my favorite song in the album in under one day! This is a significant achievement, considering the traditional process would have taken much longer. The AI technology allowed me to review the footage and adjust my prompts to convey the message I had in mind within minutes. This efficiency is a game-changer, and I'm excited to see how Runway evolves and create more videos with it. Great job Runway👏

22

2 Comments

After reading this piece about Omniretail for the umpteenth time, here are some thoughts that stand out for me (strictly within the context of the article though 😄). - The organization's approach shows a deep understanding of their market, the FMCG value chain. - They provide value to every player along the chain (through their suite of products), thus, making money off everyone ; they claim the margins are thin, so they're stacking up across the entire value chain. - Instead of trying to replace the middlemen (distributors) they partnered with them and enabled them ; a typical e-commerce play would seek to replace the middlemen. - If you read the piece, you'll notice how they claimed to have experimented, scaled and are now looking to expand. I find the information shared in the article thought provoking. You can read here 👇

4

1 Comment