Research Scientist, Robotics & World Models
Core
Designing, structuring, and evaluating data for robot foundation models (vision-language-action models, world models, manipulation and locomotion policies) to enable sim-to-real transfer and real-world policy performance.
Role type
Research Scientist (Robotics Data Science & Evaluation)
Builds
Data specifications, evaluation benchmarks, and experimental validation for robot foundation models.
Domain
Robotics, Embodied AI, Simulation, Data Engineering
Deliverable
production ML models
Required skills
Robot policy training (imitation/RL), PyTorch, dataset curation and weighting, robotics data formats (LeRobot, RLDS, Open X-Embodiment), simulation engines (NVIDIA Isaac Sim, MuJoCo), VLM fine-tuning, experimental design, academic publishing.
Preferred skills
Responsible AI evaluation, red-teaming, domain randomization, system identification.
Technologies
PyTorch, HuggingFace transformers, PEFT, NVIDIA Isaac Sim, Omniverse, MuJoCo, LeRobot, RLDS, Open X-Embodiment.
Responsibilities
Define data specifications for robot foundation models; curate and weight training mixes across real and synthetic data; guide collection across modalities (motion capture, teleoperation); build evaluation methodology for sim-to-real transfer; run experiments to validate data decisions; design adversarial evaluations; publish benchmarks and papers.
Seniority
Senior, hands-on IC
