Max Robotics — Global Service Robot Directory + US Certification

About

Haochen Shi is a roboticist at the forefront of generalist manipulation, whose work bridges large-scale imitation learning and dexterous physical interaction. His most defining contribution is to the **Open X-Embodiment** collaboration, a landmark effort that aggregated robotic data from 22 institutions into a unified dataset and trained the RT-X models. This work, amassing over 220 citations, demonstrated that diverse, cross-embodiment data can produce a single policy capable of controlling multiple robot platforms—a foundational step toward robotic foundation models. Shi also tackles the hardest problems in manipulation: soft objects and tool use. In **RoboCook**, he introduced a system for long-horizon elasto-plastic manipulation (e.g., shaping dough with a rolling pin), achieving zero-shot generalization to unseen tools. More recently, with **DexCap**, he developed a portable wrist-mounted motion capture system that collects high-quality human hand data in the wild, enabling scalable imitation learning for dexterous hands. Earlier work on **temporal emotion localization** in video further showcases his breadth in human-robot interaction. By combining large-scale data with physically grounded skill acquisition, Shi is helping to build robots that can cook, assemble, and collaborate in unstructured human environments.