Max Robotics — Global Service Robot Directory + US Certification

About

Yanjie Ze is an emerging robotics and machine learning researcher whose work sits at the intersection of visual reinforcement learning, 3D scene understanding, and robot manipulation. His research addresses fundamental challenges in enabling robots to perceive, reason about, and interact with complex real-world environments through rich visual representations. Ze's most influential contribution, "Visual Reinforcement Learning with Self-Supervised 3D Representations" (2023, 25 citations), demonstrates how incorporating 3D inductive biases into visual RL can meaningfully improve sample efficiency and generalization — a persistent bottleneck in the field. His subsequent work on GNFactor (11 citations) extended this philosophy to multi-task real-robot learning using generalizable neural feature fields, bridging 3D structural understanding with semantic reasoning. He has also pioneered reward learning from expert video through conditional diffusion models, and tackled cutting-edge embodied platforms including quadrupedal loco-manipulation, humanoid dexterous manipulation, and agile in-flight object catching. His H-InDex framework draws inspiration from human hand biomechanics to advance dexterous robotic control. Collectively, Ze's portfolio reflects a consistent drive to make robotic agents more generalizable, capable, and grounded in physically meaningful representations — positioning him as a noteworthy contributor to the next generation of intelligent robotic systems.