AI Manipulation & Control Engineer - Humanoid; Human

Your mission & challenges

You will build the intelligence that makes our hardware useful: low-level controllers, motion primitives, and the bridge to learning-based manipulation. This role spans control and machine learning: from impedance controllers and trajectory generation up to training and deploying policies from teleoperation data. You’ll work on a Jetson/Edge AI ROS 2 stack, interfacing with our real-time hardware node. You’ll also collaborate with our research partners on advanced policy development.

• Design and implement control and planning stack for high-DoF end-effectors:
  • Joint-space controllers (position/velocity/torque, impedance).
  • Motion primitives and manipulation skills.
  • Safety and constraint handling (joint limits, self-collisions, force limits).
• Define and maintain the ROS 2 interfaces for the hardware node (command, state, sensor topics).
• Use data from teleoperation pipelines to:
  • Capture large datasets of human demonstrations.
  • Train and evaluate imitation / RL or model-based policies for complex manipulation.
• Prototype and integrate perception-in-the-loop control using short-range ToF, RGB/RGB-D, and tactile sensing.
• Work with research teams to port and evaluate policies on the real hardware.
• Build evaluation benchmarks and metrics for manipulation performance and robustness.

• 3–5+ years in robotics / controls / ML (industry or applied research).
• Strong skills in Python and C++, with experience in ROS 2 and modern tooling.
• Solid background in robot dynamics & control (impedance/admittance, inverse kinematics, trajectory generation).
• Experience with at least one of: imitation learning, RL, model-predictive control, or differentiable control applied to robots.
• Hands-on experience running algorithms on real robots, not just simulation.
• Ability to work across stacks with firmware, electronics, and teleop engineers.