Research Scientist, RL for Dexterous Manipulation, Atlas
Core
Lead research on reinforcement learning for visual sim-to-real transfer and post-training of Vision-Language-Action (VLA) models to enable robust dexterous manipulation in unstructured environments.
Role type
Senior Research Scientist (RL for Dexterous Manipulation)
Builds
Novel algorithms for sim-to-real transfer, post-training recipes for VLA models, and policies for bimanual/multi-fingered systems
Domain
Robotics, Reinforcement Learning, Computer Vision
Deliverable
production ML models
Required skills
Reinforcement learning, Vision-Language-Action (VLA) models, Photorealistic rendering, Reward modeling, Offline-to-online RL, Tactile sensing, System identification, PyTorch/JAX, Large-scale model training
Preferred skills
Deployment on physical robots, Sim-to-real transfer techniques, Training pipelines combining RL/imitation/pretraining, Fine-tuning foundation models (RLHF/DPO/GRPO), Multi-fingered hand coordination
Technologies
PyTorch, JAX, VLA models, Diffusion policies
Responsibilities
Develop algorithms for visual sim-to-real transfer, Design post-training recipes for VLA models, Research reward modeling and offline-to-online RL, Close sim-2-real gap via tactile sensing and system identification, Train policies generalizing across objects and embodiments
Seniority
Senior, hands-on IC
