Santhosh Kumar

H1B Visa fees 100,000 Dollar. Time to awake indians. No other way then to get out of delusion of development. We should focus more on economy then cultural activities going on now a days. This visa row is indication of following 1. AI is really on the way to destruct and change IT sector forever. 2. Jobs placements will be lower then expected. 3. Even if this visa fees is reduced then also there will be an uncertainty on the AI changing the world order. 4. Narratives build that now india will be great again, is false i think. 5. Only way to avoid heavy losses is to truly invest in AI research.

18

In my last blog, I covered Distributed Inference Serving for LLMs. This time, I’ve written about training, breaking down DDP vs FSDP and what to use when for large-scale deep learning. Link: https://lnkd.in/gcgtSZEk #AI #DeepLearning #DistributedTraining #DDP #FSDP #PyTorch #MLTraining #Scalability #MLOps #GPUComputing #AIInfrastructure

10

1 Comment

I love seeing ML teams who refuse to treat biology as an afterthought. Relation ‘s PatchDNA work is one of the clearest examples this year of what happens when you start with how biology actually works and only then design the model. No shortcuts, no flattening of the problem just to fit the tooling. That mindset hits close to home for me. At Outpost Bio, we’re obsessed with the same question: how do you build AI that doesn’t attempt to overwrite biology, but explains it? When you’re dealing with trillions of microbes shaping what we absorb, how drugs behave, and how our bodies respond to the world (biology’s “hidden majority” that most research ignores), you can’t get away with convenient abstractions. You have to let the system speak for itself. PatchDNA is a great example of doing that well: a biologically grounded inductive bias for DNA that gives the model structure instead of asking it to guess. Huge congrats to the Relation + Rosalind teams on the NeurIPS Best Paper. This is the direction the field needs: ML that respects mechanism. Congratulations! Cristian Regep Alice Del Vecchio Lindsay Edwards Rosie Rodriguez PhD Jake P. Taylor-King Sara-Jane Dunn #NeurIPS2025 #NeurIPS 🏆 Best Paper Award: NeurIPS AI4D3 Workshop PatchDNA: A flexible and biologically informed alternative to tokenization for DNA https://lnkd.in/dPHswTGP 🎤 Talk: Saturday, December 6th, 11:30 Convention Center, Upper Level, Room 28A–E 🖼️ Poster Session: Sunday, December 7th, 15:55–17:00 AI4Science Workshop, Convention Center, Upper Level, Ballroom 20A

85

4 Comments

Bengaluru, I’m really excited to see you soon. 🇮🇳 Mid-2025, Arvind Narayanamurthy and I met through a mutual friend and instantly connected over a shared belief: AI education needs to be practical, technical, and grounded in real industry work. There was no shortage of theory or surface-level content. What felt missing were hands-on, deeply technical workshops that actually help people build agentic systems. So we ran a pilot - The Gen Academy. Over the last six months, we’ve trained close to 2,000 professionals through workshops on white-coding with AI tools, context engineering, MCP, agentic systems, RAG architectures, and frameworks like LangChain and LangGraph. The response made it clear that professionals are hungry to seriously upskill and be job-ready. Now we’re bringing this to Bengaluru. AI Launchpad is happening with Nas.io and The Hub Bengaluru, and it’s a dense, hands-on evening. Here’s what to expect: ✦ AI career and upskilling insights going into 2026 ✦ A panel with AI professionals from Google, Adobe, and more ✦ Exclusive book giveaways and resources ✦ A What’s Your Worth book signing ✦ Quizzes and nerdy hoodie giveaways The first batch of tickets sold out fast. More seats are now open. Register here: https://lnkd.in/eRR7Zfj8

314

15 Comments

Mercor, a startup connecting AI labs with domain experts for model training, has raised $350 million at a $10 billion valuation, led by Felicis Ventures V LP with participation from Benchmark , General Catalyst, and Robinhood Ventures. Initially an AI hiring platform, Mercor pivoted to supplying specialized professionals (e.g., scientists, doctors, lawyers) who train foundational AI models, earning the company hourly finder’s fees. It has since expanded into reinforcement learning infrastructure and aims to build a broader AI powered recruiting marketplace. Mercor’s growth accelerated after OpenAI and Google DeepMind reportedly cut ties with Scale AI, following Meta’s $14 billion investment in that company. The firm claims it’s on track to reach $500M ARR faster than Anysphere (Cursor). Currently, Mercor: Pays $1.5M daily to over 30,000 experts earning an average of $85/hour. Plans to focus on: Expanding its expert network Enhancing matching systems Automating internal processes Mercor’s mission is to help AI systems capture economic subtleties, like judgment, intent, and trade offs, that current models often miss. Mercor’s rise signals a powerful new layer in the AI value chain, one that bridges human expertise with machine learning at scale. As foundation models race to master nuanced reasoning and decision making, Mercor’s network of vetted domain experts provides the critical “human feedback infrastructure” these models need to move beyond surface level intelligence. This positions Mercor not just as a staffing platform, but as a strategic enabler of AI evolution, transforming expert insight into structured, monetizable training data. With $500M ARR in sight and deep adoption by top AI labs, Mercor is emerging as a potential “human capital backbone” for the next generation of foundation models, where judgment, ethics, and domain precision become as valuable as compute. #aiinvestment #aiforall #ai

17

I've said this before: cut corners smartly. Progress beats perfect planning. I still believe that. But here's what that philosophy cost us. Back in 2016, for WorkIndia, we built our data architecture around star schemas. Fluid, scalable, handled complexity beautifully. When we needed to move fast, we leaned on Django's defaults. Django's many-to-many relationships, when defined without an explicit bridge table, auto-generate one. Clean on the application side. Nobody loses sleep over it. We didn't. Then came the need for a full Data Information System. Syncing everything into a data lake via CDC pipelines. That auto-generated bridge table? No timestamp column. Timestamps are everything in CDC. If a pipeline breaks and MySQL's binlog gets purged, you cannot resume. You cannot patch. You're rebuilding from scratch. A default setting from 2016 became a crisis years later. Weeks lost. Team stretched. So what's the real lesson? Cut corners. But write down every corner you cut. Because shortcuts without a paper trail don't disappear. They just wait!

39

1 Comment

Ruokai Yin, our summer intern from Yale to Microsoft Azure, worked with Xuan Zuo, @Sattwik Deb Mishra, Hokchhay T. and Preyas Shah and me. We published a paper based on his project: https://lnkd.in/eHuG3BSt Summary of paper: Finding optimal sharding is hard especially for large LLMs. A large number of devices have to be divided into TP, PP, EP dimensions. For TP, we have to decide the tensor dimension e.g. head, hidden size to shard for every op. Additionally, TP can be one-dimensional/two-dimensional etc. This makes for a large sharding space that is hard to cover exhaustively. SoTA methods use heuristics to select sharding e.g. GSPMD, Sequence Parallelism, Helix etc. There are significant time gaps, months or years, between the discovery of each of these methods. We want to fast-forward the process of discovery by 1) using a formal representation of the sharding space, and, 2) using reinforcement learning to search the sharding space.    This paper demonstrates that our method is 3.5x better than random search, and on par with hand-tuned heuristics when looking only at 1D sharding and without allowing sequence parallelism. We aim to extend the work to multi-dimensional sharding as well as sequence parallelism. Beyond that, we wish to combine the problem of sharding with fusion, tiling and collectives as well.

58