Max Robotics — Global Service Robot Directory + US Certification

About

Yecheng Jason Ma is an emerging robotics and machine learning researcher whose work sits at the intersection of reinforcement learning, robot manipulation, and the application of large language models (LLMs) to autonomous systems. His research addresses some of the field's most persistent challenges: how to equip robots with scalable, generalizable skills without relying on costly hand-engineered solutions. Ma's most recognized contribution, the DROID dataset (108 citations), represents a landmark effort in democratizing robot learning by providing a large-scale, in-the-wild manipulation dataset that supports more robust policy training. His Eureka framework (48 citations) broke new ground by demonstrating that LLMs can autonomously design reward functions capable of teaching robots complex dexterous behaviors like pen spinning — a task previously beyond automated reward specification. Complementing this, his VIP and LIV frameworks (35 and 24 citations respectively) advance reward and representation learning directly from human video and language supervision, reducing dependence on task-specific robot data. Ma has also contributed to safe reinforcement learning and offline RL, reflecting a broad commitment to making robot learning both capable and reliable. Across his body of work, his research consistently pushes toward systems that learn richer skills with less human intervention — a vision central to the future of general-purpose robotics.