Suvansh Sanjeev​

I'm Suvansh, a researcher at OpenAI where I led a reinforcement learning team that co-created algorithms shaping post-training for every model release from GPT-5 onwards. I contributed to GPT-4.1, GPT-5, and o1, and helped revive OpenAI's robotics program by training its first vision-language-action models. Before OpenAI, I did my MS in Robotics at CMU and researched deep RL at Berkeley under Sergey Levine. Ask me about reinforcement learning for post-training, reasoning models, robotics and VLAs, and AI research at frontier labs.