Sergey LevineSergey Levine

UC Berkeley professor who developed end-to-end robot learning from raw pixels. His work demonstrated that robots could learn complex manipulation tasks directly from visual input using deep reinforcement learning, without hand-engineered features. Key contributor to offline RL (learning from pre-collected datasets rather than online interaction). Co-founded Physical Intelligence (pi), a company building foundation models for robotics. His research bridges the gap between the success of large language models and the physical world.