MLOps Engineer

AIRoA (AI Robot Association) runs a national initiative collecting one million hours of humanoid robot operation data to train large-scale Vision-Language-Action (VLA) models, releasing datasets, models, and benchmarks openly. This role designs and maintains large-scale ML pipelines and distributed training clusters, optimizing model performance and shortening development cycles. It works closely with VLA researchers on ML infrastructure, data pre-processing, training, monitoring, evaluation, and deployment, optimizes infrastructure for cost, performance, and reliability, and builds MLOps tools for researchers to visualize and analyze performance. Required: a Master's degree in Computer Science, Engineering, or a related field (or equivalent practical experience), 5+ years of professional experience in MLOps engineering, machine learning, or robotics, experience deploying distributed systems on AWS, GCP, or Azure, experience with orchestration tools such as Airflow, Dagster, or Kedro, and high proficiency in PyTorch or JAX. Preferred: experience fine-tuning RL, VLM, and VLA models, ETL/ELT pipelines, hyperparameter optimization, distributed training/cluster management, GPU memory optimization, robotics sensor data processing, and ROS/ROS2. Business-level English is appreciated but not required. Full-time, based in Heiwajima, Japan.