Robotics Software Engineer

Description

The team is building software that lets robots learn dexterous manipulation from human data, working on real hardware with dexterous hands and grippers.

The work spans control, perception, teleoperation, simulation, and deploying learned policies, taking research ideas to working systems on physical robots.

No prior experience with robot foundation models is required, but you should be keen to apply modern machine learning to robotics.

Responsibilities

• Write and maintain software for robot arms, dexterous hands, and grippers, covering control, sensor integration, and real-time execution.
• Build teleoperation, motion-retargeting, and data-collection systems that map human hand motion, force, and muscle-activity data onto dexterous hands and grippers with differing kinematics.
• Create perception for the robots and capture rigs, including multi-camera calibration and synchronization, hand and object tracking, and 3D state estimation for teleoperation and policy deployment.
• Connect cameras, tactile sensors, force/torque sensors, and other sensing types into robotic systems.
• Deploy learned manipulation policies on physical robots and evaluate them, finding the source of failures in perception, planning, control, or hardware.
• Build simulation environments and evaluation tools for robot learning experiments.
• Collaborate with machine learning and biomechanics researchers on imitation learning, reinforcement learning, and other dexterous manipulation methods.
• Build the runtime, data, and evaluation tooling that the rest of the team uses to run experiments quickly.

Requirements

• Strong software engineering skills in Python, C++, or both.
• Experience building software for physical robots, simulators, or closely related systems.
• Solid grounding in robotics fundamentals such as kinematics, dynamics, motion planning, control, or state estimation.
• Ability to debug complex systems that span hardware and software without close supervision.
• Experience with dexterous hands, contact-rich manipulation, or force and impedance control is a plus.
• Experience with computer vision, 3D perception, camera calibration, or pose estimation is a plus.
• Experience with ROS/ROS2, robot simulators, or robotics frameworks is a plus.
• Experience with robot learning, imitation learning, reinforcement learning, or PyTorch is a plus.
• 3+ years of experience.

Conditions

Full-time, on-site in San Francisco.

Visa sponsorship is available.

The company is a small research and engineering startup developing ways for robots to learn dexterous skills from human biomechanics. Its founders have AI research backgrounds at MIT and NVIDIA, plus autonomous-vehicle and robotics deployment experience at Pony ai and Dexterity.

You will work directly with the founders, own problems end to end, and test ideas on real robotic systems.