Max Robotics — Global Service Robot Directory + US Certification

About

Ryan R. Julian is a leading researcher at the intersection of robotics, natural language processing, and reinforcement learning, whose work is defining how robots learn to interact with the physical world. He is best known for pioneering the integration of large language models (LLMs) with robotic control, most notably through his highly influential paper "Do As I Can, Not As I Say" (over 500 citations), which introduced a framework for grounding semantic knowledge from LLMs into robotic affordances. This work fundamentally shifted the field by enabling robots to follow high-level, temporally extended instructions without requiring explicit task programming. Julian also led the development of RT-1, the Robotics Transformer, which demonstrated how large-scale, task-agnostic datasets could be leveraged for real-world control. His practical impact is further evidenced by his work on deploying deep reinforcement learning at scale, including a system for sorting waste in office buildings using a fleet of mobile manipulators—a landmark achievement in real-world RL. Most recently, he contributed to the Gemini Robotics family of models, designed to bridge the gap between generalist AI and physical agents. Julian’s research consistently pushes the boundaries of what robots can achieve autonomously in unstructured environments.