Shubham Sharma is the Founder and CEO of Kailash Labs, a Y Combinator Fall 2026 company building small, specialized video reasoning models that turn video into structured data.

Shubham studied Aerospace Engineering at IIT Bombay with a minor in AI, and was a visiting researcher at KAIST working on robot learning and long-term memory for decision-making agents.

He published a NeurIPS 2022 spotlight paper on offline reinforcement learning during his research at Tsinghua. He is driven by one question: how to unlock the insights buried in the petabytes of video humans generate every day.

Kailash Labs is part of Y Combinator's Fall 2026 batch, and PitchBook lists a $500K accelerator investment from Y Combinator. Its approach takes video plus a customer's objective, distills a frontier model, and post-trains a small open model for that single task.

Its first product, Marlin, is a 2B-parameter open-weights video model that matches Gemini-2.5-flash on dense captioning and temporal grounding at a fraction of the cost. The small models can also be deployed on-prem where sensitive video lives.