Research Scientist

The AI Robot Association (AIRoA) is building an open robot data ecosystem, collecting one million hours of humanoid robot operation data with hundreds of robots to train Vision-Language-Action (VLA) models. This research scientist role is responsible for training and evaluating those models.

The scientist designs and implements data preprocessing pipelines for multimodal robot datasets and trains VLA models with supervised learning, reinforcement learning, fine-tuning, RLHF and training from scratch. The models are developed and evaluated in simulation and on physical robots. The scientist also improves training efficiency, analyzes results and publishes at top conferences such as NeurIPS and CoRL.

Applicants need an MS with at least three years of industry experience, or a PhD, in computer science, electrical engineering or a related field, plus at least one first-author paper at a top venue. Required experience includes reinforcement learning and RLHF or reward design for LLMs, VLMs or VLAs, work with simulators or real robots, and PyTorch. Preferred are robot learning research, tactile sensing, MuJoCo or Isaac Sim, and ROS.