Software Engineer - Manipulation

Description

The company builds AI-powered humanoid robots designed to work alongside people, starting with manufacturing and logistics and later extending to healthcare and home settings.

The team works across the full robotics stack and is focused on taking the robot to market at scale, including safety, commercialization, and mass production.

This role shapes how the robot handles objects reliably and precisely. You will build and deploy learning-based manipulation policies, with a focus on post-training methods that refine those policies on real tasks and physical hardware.

You will turn research ideas into production software that runs across the robot fleet, moving policies from demonstration data through post-training to deployment, and working closely with the simulation, data, hardware, and systems teams.

Responsibilities

• Implement and deploy state-of-the-art learning-based manipulation policies that perform reliably on real tasks with physical hardware.
• Adapt and improve policies for new tasks and hardware using RL fine-tuning, policy distillation, and interactive imitation learning such as DAgger.
• Build and maintain data pipelines for post-training, covering teleoperation, demonstration collection, retargeting, dataset aggregation, and curation.
• Lead the cycle from simulation and offline evaluation to transferring and fine-tuning policies on the robot.
• Develop visuomotor policies for grasping, in-hand manipulation, and whole-arm manipulation that execute reliably with low latency on hardware.
• Define evaluation metrics and regression tests for policies in simulation and on hardware, and use the results to inform technical decisions.
• Work with the simulation, data, hardware, and systems teams to diagnose system-level issues and build solutions together.
• Keep the codebase well tested and documented, and review peers' code.

Requirements

• Must-have: Strong experience in robotic manipulation, including grasping, in-hand or whole-arm control, and high-DOF motion, with a record of deploying on hardware.
• Must-have: Hands-on experience training and debugging learned policies, including post-training methods such as RL fine-tuning, distillation, and imitation learning like DAgger.
• Must-have: Proficiency in Python, C++, and PyTorch or JAX, with experience building real-time robotic software stacks.
• Must-have: Experience with demonstration and teleoperation data and with building evaluation pipelines for learned policies.
• Good to have: Experience with physics engines such as IsaacSim, MuJoCo, or Drake for training and validating policies.
• Good to have: Experience adapting large general-purpose manipulation policies to specific tasks and embodiments.
• Good to have: Experience with VR or haptic teleoperation interfaces and retargeting algorithms for human-in-the-loop control.
• Good to have: Experience integrating tactile or haptic feedback into manipulation pipelines.
• Good to have: Familiarity with 6D pose estimation, point cloud processing, or visual servoing.
• An MS with 3+ years of relevant experience, or a recent PhD, in robotics, computer science, electrical engineering, or a related field.
• Relevant experience focused on learning-based manipulation or a closely related area.
• A proven record of moving learned policies from research or simulation into successful deployment on physical hardware.

Conditions

Physical demands include long periods of sitting at a desk and working on a computer, and occasionally lifting 15 pounds.

The role requires good vision to read printed material and a computer screen, and the ability to hear and speak to communicate.

The annual salary range is $180,000 to $240,000.

This is a direct hire position, and agency submissions are not accepted.

Equal opportunity: applicants are considered without regard to protected characteristics.