Max Robotics — Global Service Robot Directory + US Certification

About

Jan Humplik is a robotics researcher whose work sits at the intersection of deep reinforcement learning, sim-to-real transfer, and agile locomotion for legged and humanoid robots. His most celebrated contribution — teaching a low-cost miniature humanoid robot to play one-versus-one soccer using deep RL — has garnered over 147 citations and stands as a landmark demonstration that sophisticated, compositional behaviors can emerge from learned movement primitives in real hardware. This line of work showcases his talent for bridging the notoriously difficult gap between simulation and physical deployment. Humplik has also advanced the field through creative use of neural radiance fields for realistic sim-to-real visual transfer (NeRF2Real, 43 citations) and explored how large language models can be leveraged to synthesize robotic reward functions from natural language instructions (Language to Rewards, 38 citations). His research on imitating human and animal motion capture data to build reusable locomotion skills further reflects his broad methodological range. Most recently, Humplik contributed to Google DeepMind's ambitious Gemini Robotics initiative, signaling his involvement in frontier efforts to bring general-purpose AI into physical agents. Across his body of work, he has consistently pushed the boundaries of what agile, vision-guided robots can achieve in unstructured, real-world environments.