Vignav Ramesh

I work on pretraining research at OpenAI (on leave from Harvard). Previously at OAI, I led a team across research, product, and deployments working on 0→1 bets (and before that, scaled test-time compute with the reasoning team). Broadly, my academic interests center around sample-efficient learning, self-play, and multi-agent systems.

Before OAI, I led efforts on these topics in both industry (NVIDIA Research) and academia (MIT CSAIL, Harvard’s Kempner Institute, Stanford NLP; 3 patents, 12+ papers at NeurIPS, ICML, Nature, etc). Previously, I built Arc, a tooling layer for enterprises to provision teams of specialized LLM agents (deferred seed to continue undergrad); and headed research at startups having raised $100M+ from the likes of OpenAI & a16z.