Careers

Your work will include

• Taking customer tasks from first site visit to a robot that runs reliably in production.
• Building core systems across training, simulation, evaluation, deployment, and monitoring.
• Training and adapting robotic control models, and building the data pipelines they depend on.
• Integrating hardware, sensors, and robot software into complete, maintainable applications.
• Helping recruit, interview, and onboard the engineers who join after you.

We’re looking for someone who

• Has strong software engineering fundamentals and a record of shipping substantial systems end to end.
• Brings depth in at least one of robotics, robot learning, simulation, controls, or ML infrastructure, and is willing to work outside it.
• Makes sound decisions with incomplete information and revisits them when the evidence changes.
• Takes ownership of outcomes.
• Communicates clearly.
• Wants the responsibility and the ambiguity that come with being early.

Previous experience at an early-stage startup, deploying robots or automation in industrial setting would be useful. We value demonstrated ability.

Success means customers have robots doing real work on their lines, and the company has the technical foundation to do that repeatedly and faster each time.

Your work will include

• Building dependable infrastructure for running, scaling, and reproducing robotic training workloads.
• Developing simulation environments, reusable assets, physics models, and sensor models.
• Building interfaces between simulation, training frameworks, robot software, and physical hardware.
• Investigating discrepancies between simulated and physical behavior and improving fidelity where it affects learning and deployment.

We’re looking for someone who

• Has 3+ years of experience
• Has strong software engineering fundamentals and can build, debug, and maintain substantial systems.
• Brings depth in simulation, machine learning infrastructure, robotics software, distributed systems, or controls.

Experience with robotics simulation.

Success means engineers can use your infrastructure to evaluate and train robotic systems efficiently

Your work will include

• Understanding customer workflows and translating them into task definitions, operating requirements, and measurable success criteria.
• Selecting hardware, sensors, and tooling appropriate to the task and its constraints.
• Connecting results to the data, software, model, hardware, and configuration versions that produced them.

We’re looking for someone who

• Has experience developing, training, and evaluating models for robotics or another demanding applied setting.
• Investigates failures systematically across models, software, hardware, data, and workflow.
• Can make progress with incomplete requirements while keeping scope and success criteria explicit.
• Is comfortable working directly with physical systems and at customer sites.
• Follows through until the customer’s outcome is achieved and verified.
• Communicates clearly

Experience in robot learning, perception, controls, industrial automation would be useful.

Success means customers achieve their agreed outcomes and continue achieving them after commissioning.

Your work will include

• Training and adapting robotic control models for specific customer tasks, including imitation learning, reinforcement learning, and post-training of pretrained vision-language-action models.
• Designing data strategies: deciding what demonstrations, teleoperation data, simulated rollouts, or synthetic data a task needs, and how much.
• Diagnosing model failures from evaluation and deployment data, and determining whether the fix belongs in the data, the objective, the architecture, or the task definition.
• Turning successful approaches into reusable training recipes.
• Following relevant research and judging which ideas are worth testing against our problems.

We’re looking for someone who

• Has 2+ years of experience
• Has trained machine learning models for robotics.
• Has strong engineering skills in Python and a deep learning framework such as PyTorch or JAX, and writes research code that others can run and build on.
• Understands modern robot learning methods well enough to choose among them, and to recognize when a simpler approach will do.
• Designs experiments that isolate cause and effect, and is skeptical of results that look better than they should.
• Reasons about the physical side of learning, including sensing, latency, contact, calibration, and hardware variation, and how each shows up in model behavior.
• Communicates clearly

Experience with robotics models. Published work in robot learning would be useful. We value demonstrated results on physical systems alongside publications.

Success means our models reach the reliability customers need.