Intern Research Scientist

AIRoA (The AI Robot Association) is building a large-scale humanoid robot data initiative, collecting data from hundreds of robots to train Vision-Language-Action (VLA) models, with datasets, models, and benchmarks made open. This part-time Intern Research Scientist role is based in Heiwajima, Japan, within the Development Division, on a project backed by a 20.5 billion yen investment from NEDO. Responsibilities include designing data preprocessing pipelines for multimodal robot datasets; training VLA models via supervised learning, reinforcement learning, fine-tuning, RLHF, and training from scratch; developing and evaluating models in simulation and on physical robots; improving training robustness and efficiency; analyzing model performance; deploying VLA models onto real humanoid and mobile robots; and publishing research at conferences such as NeurIPS, CoRL, and CVPR. Required: an MS or PhD in Computer Science, Electrical Engineering, or a related field; a publication record at leading ML, robotics, or computer vision conferences (NeurIPS, ICML, ICLR, ICRA, IROS, RSS, CoRL, CVPR, ICCV, ECCV); experience with reinforcement learning and frontier LLM/VLM/VLA training methods such as RLHF; experience with simulators or real-world robots; and experience with PyTorch. Preferred: PhD-level robot learning research, hardware control experience, tactile sensing for manipulation, simulation platforms like MuJoCo or Isaac Sim, and ROS experience.