About the Role
We are looking for an engineer who will own the full locomotion and whole-body control stack for our humanoid robot. Your work will directly shape the robot’s ability to walk, balance, and move safely and smoothly in the real world.
This is a highly hands-on role. You will spend your time building training pipelines in simulation, iterating on RL and IL policies, transferring them to hardware, and debugging behavior on real robots.
What You’ll Do
Develop, train, and evaluate RL and IL policies for whole-body control and locomotion.
Build and optimize training pipelines in Isaac Lab and real-time inference pipelines on robot hardware.
• Drive sim-to-real transfer , including domain randomization, curriculum design, and iterative policy refinement.
• Work directly with hardware to diagnose failures, tune controllers, gather datasets, and improve stability and performance.
• Own experiments end-to-end: from idea → prototype → simulation → real robot deployment.
• Collaborate on motion, control, perception, and high-level planning systems as we scale capabilities.
Ideal Background
We’re looking for someone who is both a strong RL engineer and a practical roboticist—someone who enjoys seeing their work running on a real machine, not just in a paper or a simulation.
You likely have experience in:
• Reinforcement Learning
• Isaac Lab , Isaac Gym, MuJoCo, or similar physics simulators.
• Building training pipelines with PyTorch .
• Deploying policies on embedded or GPU-accelerated systems (C++/Python/JAX/etc).
• Whole-body control, locomotion control, or quadruped/humanoid robotics.
• Working with real robots — debugging hardware, evaluating behavior, collecting rollouts.
Bonus experience (not required):
• Unitree or similar humanoid/quadruped platforms.
• ACT, Diffusion Policy, GR00T, or other imitation learning methods.
• Motion planning, model predictive control, or low-level torque control.
Who Thrives Here
You're a great fit if you:
• Prefer real results over academic elegance .
• Love tuning, tweaking, and iterating rapidly.
• Are excited to push a robot until it breaks—and then fix it.
• Are comfortable owning a large scope and moving fast with incomplete information.
• Get deep satisfaction from seeing something you built controlling a physical system.
This role is not a fit if you primarily want to publish papers, or work in large slow-moving orgs..
What This Role Offers
• Significant ownership over a foundational part of the robot’s capabilities.
• The ability to ship work directly to hardware from day one.
• A seat on the ground floor of an ambitious robotics team.
• Fast iteration cycles, huge autonomy, and the chance to define core technical systems.