Research Scientist at Emergence AI, working on reinforcement learning with verifiable rewards for agentic systems.
I build certificates and safety filters that let you deploy a learned policy without having to trust it: reachability and barrier certificates, offline safe RL, and safety for VLA and agentic systems. Published at ICML, TMLR, RLC, CDC and ACC.
PhD from the Centre for Cyber Physical Systems, IISc Bangalore, advised by Prof. Shishir N. Y. Kolathaya and Prof. Pushpak Jagtap. B.Tech from IIT Bombay.
Recent work with code
- V-OCBF (TMLR 2026) — learning safety filters from offline data
- Safe Flow Q-Learning (RLC 2026) — offline safe RL with flow policies
- PIML-SOC (ICML 2025) — physics-informed safe and optimal control
- Vision CBF (CDC 2025) — Vision Based CBF for autonomous systems using World Models




