Research Scientist - Robot Learning (VLA / WAM)
Core
Train end-to-end vision-language-action (VLA) and world-action models (WAM) to enable robots to act in physically-grounded 3D environments.
Role type
Senior hands-on research scientist (robot learning / embodied AI)
Builds
Robot policies that close the sim-to-real gap for manipulation and interaction
Domain
Robotics, Generative AI, Computer Vision, Simulation
Deliverable
production ML models
Required skills
Robot policy design (VLA, WAM, diffusion), Imitation learning, RL fine-tuning, VLM adaptation for control, Action representation design, Distributed training (FSDP), Python, PyTorch
Preferred skills
Domain randomization, System identification, Calibration, Tokenization, Chunking, Diffusion, Flow-matching, Supervised fine-tuning
Technologies
PyTorch, FSDP, VLM backbones, Diffusion models, Flow-matching
Responsibilities
Own end-to-end training pipelines from data to robot deployment; Set technical direction for embodied research; Close sim-to-real gap via domain randomization and calibration; Adapt VLM backbones for control; Curate heterogeneous robot datasets; Design action representation and decoding; Run post-training (SFT and RL)
Seniority
Senior, hands-on IC
