Learning systems
for a world in motion.

Explore predictive models that connect multimodal observation, action, and consequence. The work would span representation learning, generative modeling, uncertainty, planning, and transfer across tasks and embodiments.

Turn learning ideas into reliable training systems for imitation, reinforcement, and self-supervised learning. This role would bridge fast research iteration with reproducible experiments at scale.

Build the on-robot systems that connect sensing, inference, planning, and control. The focus would be robust real-time behavior, calibration, observability, and safe deployment across platforms.

Design the compute, experiment, and evaluation systems behind rapid model development. The work would make large training runs observable, repeatable, and easy for a small research team to operate.

Build pipelines for ingesting, curating, versioning, and inspecting multimodal robot experience. The goal would be trusted datasets with clear provenance, measurable quality, and fast paths back into training.

Create scenario suites and evaluation tools that reveal how systems fail, generalize, and improve. This role would connect simulation, controlled lab testing, and real-world evidence into one measurement loop.

Keep robot experiments productive in the lab and the field. The work would combine platform maintenance, test protocol design, data capture, incident analysis, and disciplined feedback to research and engineering.

Translate research needs into precise data programs with external partners. This role would shape collection specifications, vendor workflows, quality thresholds, delivery milestones, and responsible data practices.