Yash PatilYash Patil

Former OpenAI researcher who worked on Codex, then co-founded Applied Compute in 2025 with fellow ex-OpenAI researchers Rhythm Garg and Linden Li. The company post-trains and serves open-weight models on an enterprise's own data with reinforcement learning — betting that a specialised model owned by the customer beats renting a frontier API. Raised $80M in April 2026 at a $1.3B post-money valuation, and in September 2026 told Forbes it was in talks to raise $350M at $3.25B, with roughly $50M in annualised revenue. Argues post-training is where inference economics are now won.