Senior Member of Technical Staff: Reinforcement Learning for Wholebody Control

About the Role

You will develop the learned control policies that give our humanoid robots robust, capable whole-body motion—and get them running on real hardware in real deployments. This role sits at the intersection of reinforcement learning and classical control: you'll train policies in large-scale simulation, blend them with model-based control where that's the right tool, and own the sim-to-real pipeline that turns a promising policy into dependable behavior on a physical robot. Your success is measured by robots that stand, balance, and manipulate reliably in the field, and by a training-and-deployment pipeline that lets the team ship improved policies again and again.

Responsibilities

• Design, train, and tune reinforcement learning policies for whole-body control—balance, locomotion, and coordinated manipulation—on a humanoid platform.
• Combine learned policies with classical and model-based control techniques, choosing the right blend for robustness, safety, and performance. Interface the control stack smoothly with high-level motion commands.
• Own the pipeline that transfers policies from simulation to hardware—domain randomization, system identification, and the iterative loop of closing the sim-to-real gap on real robots.
• Build and maintain the infrastructure to train, evaluate, version, and deploy policies repeatably, so improvements reach the fleet reliably and safely.
• Bring policies up on physical robots, debug behavior in the real world, and harden them against the variability of real environments.
• Develop rigorous evaluation—in sim and on hardware—and ensure learned controllers behave safely and degrade gracefully at their limits.
• Partner with controls, hardware, simulation, and AI engineers to define interfaces, improve models, and integrate policies into the broader robot stack.

Requirements

• Proven experience developing and training RL policies for continuous control, with strong command of modern RL algorithms and their practical failure modes.
• Demonstrated success transferring learned policies from simulation to physical robots and deploying them on real hardware.
• Solid grounding in control theory and robot dynamics, and the judgment to combine learned and model-based approaches.
• Hands-on experience training policies in GPU-accelerated simulation at scale.
• Strong Python and working C++ skills, with the ability to build training and deployment infrastructure that others can rely on.
• Experience with whole-body control, locomotion, or manipulation on legged, humanoid, or similarly high-DOF robots.
• Track record of taking policies from concept through training, transfer, and field deployment.

Preferred Qualifications

• Experience with model predictive control, whole-body QP controllers, or trajectory optimization.
• Background in system identification, actuator modeling, or contact-rich dynamics.
• Experience building ML training infrastructure and experiment-tracking workflows at scale.
• Familiarity with imitation learning, teleoperation data, or learning from demonstration.
• Publications or demonstrated results in legged locomotion or whole-body control.

Benefits

• Competitive total compensation program, including salary, annual cash bonus, and company equity.
• Company-subsidized insurance programs.
• 401(k) with company match.
• Flexible PTO.
• Daily lunch.