Humanoid Whole-Body Control: Build and improve software for large-scale motion tracking and loco-manipulation in simulation.
Policy Deployment on Robot Hardware: Deploy learned whole-body control policies on humanoid platforms and optimize the runtime stack for low-latency, reliable execution.
Real-Time Kinematic Motion Planning: Develop planners and interfaces that convert task-level inputs, joystick commands, teleoperation signals, human motion, or VLA outputs into motion targets that a whole-body policy can track.
Simulation-to-Real system-id: Work across simulation, hardware-in-the-loop testing, and physical robot experiments to diagnose and close the gap between policy behavior in simulation and behavior on hardware.
Performance and Reliability Engineering: Profile and optimize inference, control-loop timing, data flow, GPU utilization, and robot-side runtime performance.
Robot Testing and Debugging: Debug failures across the full stack: motion representation, state estimation, policy inference, robot model mismatch, actuator limits, contact behavior, latency, and hardware safety constraints.
What we need to see:
A PhD in Robotics, Machine Learning, Computer Science, Electrical Engineering, Mechanical Engineering, or a related field (or equivalent experience) with at least 3 years of research and engineering experience.
Reinforcement Learning for Control: Strong experience with reinforcement learning for robotics, including policy training, reward design, motion tracking, curriculum learning, domain randomization, and sim-to-real deployment.
Simulation and Synthetic Training Pipelines: Hands-on experience building and scaling robot simulation environments in Isaac Lab or similar platforms. You should be comfortable debugging physics, contacts, sensors, robot models, and large-scale training workflows.
Whole-Body Control and Motion Tracking: Understanding of humanoid or legged robot control.
Systems Programming: Familiar with C++ for real-time robotics systems and strong Python for training, simulation, tooling, and experimentation.
Mathematical and ML Foundations: Understanding of rigid-body dynamics, neural network architectures, and modern learning-based control methods.
Additional Information
We are building the behavior foundation models for humanoid robots. As a Full-Stack Engineer for Humanoid Whole-Body Control and Loco-manipulation, you will help train large-scale controllers and use the controller as a reliable behavior foundation on real humanoids. You will work across simulation, policy deployment, kinematic planning, and robot testing to ensure that high-level motion commands become stable, expressive, and physically feasible whole-body movement.
Nvidia · Shanghai, China