About the Role

Join a physical AI data and infrastructure team as its first robot learning hire. You will train and evaluate robot policies using synthetic data, establish whether that data improves performance, and turn findings into public results and product requirements.

What You'll Do

Train robot policies using VLA fine-tuning, imitation learning, diffusion policies, or reinforcement learning in simulation, focusing initially on manipulation and grasping.

Establish baselines on public evaluation suites and compare results with and without synthetic training data using controlled protocols.

Report improvements and failures, then translate findings into data requirements covering generation, fidelity, and variation.

Publish and demonstrate results to the robotics community and participate in customer technical discussions.

Begin as an individual contributor and help hire and lead a growing Bay Area robotics team.

What We're Looking For

Six or more years of hands-on robot learning experience, including work in industry and a university research lab.

A graduate degree in robotics, machine learning, or a related field.

Experience training and evaluating robot policies in simulation with Isaac Sim, Isaac Lab, MuJoCo, or equivalent tools.

Practical experience with methods such as VLA fine-tuning, imitation learning, diffusion policy, ACT, behavior cloning, or reinforcement learning.

Experience with manipulation or grasping policies, and the ability to run controlled comparisons with fixed seeds, checkpoints, and evaluation suites.

Strong Python and PyTorch skills, with familiarity with NVIDIA Isaac and USD; experience using evaluation results to guide data or simulation changes.

Compensation & Benefits

Base salary: $260,000 to $300,000 annually. Visa sponsorship is not available.

Location

Hybrid in San Francisco, California, United States.