Kevin Bergamin is a Senior Staff Research Scientist at Boston Dynamics. He holds a Master’s degree in Mechanical Engineering from McGill University, where his thesis and SIGGRAPH publications focused on applying RL to humanoid control in simulation. He also has experience from the AAA games industry, specializing in computer graphics, animation, and simulation.
Since joining Boston Dynamics in 2021, Kevin has worked on a wide range of projects. His contributions include developing robust locomotion on Spot using hybrid MPC and RL controllers, learning manipulation skills on Atlas from external human video demonstrations, motion capture retargeting methods for whole-body imitation learning on Atlas, and retargeting human hand motion to non-anthropomorphic robotic grippers in both simulation and hardware. Over the past year, his primary focus has been the Large Behavior Model collaboration with TRI, where he led the development of whole-body teleoperation using MPC, along with tools and software for large-scale data collection and policy deployment on robot hardware and in simulation.
Before Boston Dynamics, Kevin worked at Ubisoft Montreal’s La Forge R&D lab, where he developed physically based animation techniques and tools for humanoid characters. His work combined reinforcement learning with large motion capture datasets to produce high-fidelity, responsive character controllers. He developed new techniques using Motion Matching, creating some of the first highly steerable RL locomotion policies for humanoids trained with imitation learning in 2019.
Large Behavior Models and Atlas Find New Footing