Max Robotics — Global Service Robot Directory + US Certification

About

Jon Scholz is a leading researcher in robot learning, whose work is driving the field toward generalist agents capable of rapid, autonomous adaptation. His primary research areas include meta-reinforcement learning, few-shot visual imitation, and sim-to-real transfer for complex manipulation. Scholz’s most influential contribution is his work on offline meta-RL for industrial insertion (66 citations), which demonstrated how robots can leverage past experience to master new tasks with dramatically fewer trials. He further advanced this vision with RoboTAP (31 citations), a method for tracking arbitrary visual points that enables few-shot imitation learning without task-specific engineering. As a core contributor to the RoboCat project, Scholz helped develop a self-improving generalist agent that can quickly learn novel skills across different robot embodiments. His more recent work on DemoStart introduces a demonstration-led auto-curriculum for sim-to-real transfer, enabling complex multi-fingered manipulation from sparse rewards. Across these projects, Scholz has consistently tackled the fundamental challenge of data efficiency, pushing robots closer to the flexibility and generality needed for real-world applications beyond structured factory settings.