VLA Pre-training Engineer (Deep Learning)
Core
Pre-train and fine-tune Vision-Language-Action (VLA) policies for humanoid robots using deep learning, managing the full loop from data curation to deployment.
Role type
Senior IC deep learning engineer (robotics/VLA)
Builds
Capable robot policies and VLA models for industrial humanoid platforms
Domain
Robotics + Deep Learning (LLMs/VLMs)
Deliverable
production ML models
Required skills
Deep learning system building, LLM/VLM/generative model architecture and training, distributed training strategies, Python, PyTorch/JAX, data pipeline engineering, MLOps
Preferred skills
Robotics or autonomous driving experience, RL applied to LLMs/robotics, VLA model experience, on-device optimization, open-source VLA framework familiarity
Technologies
PyTorch, JAX, OpenVLA, Physical Intelligence (π) models
Responsibilities
Post-train policies via behavior cloning and RL; partner with Data Collection to drive data gathering; run pre-/mid-/post-training on VLA stack; build continuous pipelines for synthetic data and teleop logs; scale distributed training and optimize models for real-time edge inference
Seniority
Senior, hands-on IC