I work on the data that shapes Gemini. I write about the agents that will replace it.
Inside the Evolution Arena and the rise of autoresearch.
An AI agent that proposes its own experiments, runs them against a live 2D simulation, scores the result with a hard mechanical metric, and commits or reverts on its own. A ratchet that only moves forward.
How particle swarm optimization escapes the local-minima trap in ARC-AGI.
Standard LLM agents get stuck. The fix: a PSO-governed swarm of specialized LLM particles with a continuous fitness function that rewards near-misses. The swarm provides strategy; the LLM provides syntax.
Local Tinker — a clean API for local LLM fine-tuning.
A Tinker-style API for LoRA fine-tuning of 1B–13B LLMs on your own GPU. Four primitives cover SFT, DPO, PPO, and GRPO — without the HuggingFace + PEFT + bitsandbytes boilerplate.
A visual encyclopedia of the ideas behind modern ML.
Editorial playgrounds, one concept at a time — starting with the linear-algebra foundations (matrix transformations, eigenvectors, PCA, SVD) and working up through gradient descent, regression, neural networks, attention, diffusion, and beyond.
Backend storage and data integrity for the engine generating RLHF and SFT datasets that train Google's foundation models. Plus a low-latency LLMOps platform for prompt versioning at enterprise scale.
Architected and shipped a centralized ingestion platform processing 10B+ events/day on Kinesis, Lambda, and S3. Resilient by default; the data lake's front door.
Master's thesis on multi-source odor localization with swarm robotics — the work behind US Patent 8,838,271 for nuclear spill detection. Cited 12× and the conceptual scaffolding behind the multi-agent reasoning work I'm doing now.
Multi-agent systems for the ARC-AGI challenge — abstract reasoning frontier.
Swarm-based search applied to ARC-AGI — bio-inspired optimization for reasoning.
RSI framework for LLMs — iterative capability enhancement.
Deep Q-Network agent that learns to play Snake from raw reward.
I started as an aerospace engineer designing intake systems for missiles, became a swarm-robotics researcher under Debasish Ghose at IISc, picked up a US patent and a Cornell MEng on the way, and have spent the last decade shipping data infrastructure at Amazon and Google.
Today I work on the data engine behind Google's foundation models. Off-hours, I write essays and run experiments on multi-agent reasoning — the same swarm research continued in a new vocabulary.
Saratoga, California · Senior Member, IEEE · FBCS
Also: Senior Member of IEEE · FBCS · Hackathon Raptors Fellow · IEEEXtreme Proctor · GATE 99th-percentile scholar. Mentor at Google Tech Exchange, Founder Institute, and Opportunity Machine.