Software Engineer - Dexterous Manipulation

Job description

Job responsibilities

• Implement and tune control algorithms for multi-fingered hands—grasping, in-hand manipulation, and tactile-feedback integration.
• Develop and maintain manipulation software with low-latency, reliable execution inside the real-time controls stack.
• Translate state-of-the-art methods (RL / imitation policies, human-to-robot motion retargeting) into production-grade C++/Python.
• Build and refine sim-to-real pipelines—hand models, domain randomization, and validation in IsaacSim / MuJoCo / Drake.
• Deploy and debug manipulation capabilities on physical robots; diagnose sensor-noise, latency, and calibration issues.
• Collaborate with hardware and systems engineers, feeding back on hand performance and sensor/actuator requirements.
• Uphold code quality through rigorous testing, documentation, and peer review.

Job requirements

• Dexterous manipulation experience: multi-fingered grasping, in-hand manipulation, and high-DOF hand control.
• Reinforcement learning for robotic control—reward design, training, and debugging—ideally applied to dexterous manipulation (imitation learning and diffusion policies a plus).
• Strong Python and working C++ for real-time robotic software.
• Hands-on experience training and validating policies in a physics simulator (IsaacSim, MuJoCo, or Drake).
• Robotics fundamentals: kinematics, dynamics, and Jacobian-based control.
• A track record of getting algorithms working on physical hardware, not simulation alone.
• Nice to have: teleoperation (VR/haptic) and retargeting, tactile-sensing integration, computer vision (6D pose / point clouds), and end-effector bring-up & calibration.
• BS/MS/PhD in Robotics, Computer Science, Electrical Engineering, or a related field.