AI training work

Robotics & Physical-AI Data

Generate and annotate the data that teaches robots and embodied AI to act in the physical world: label robot and egocentric video, annotate vision-language-action (VLA) demonstrations, collect teleoperation and manipulation data, and quality-check robot-perception datasets. This is the newest frontier of AI-training work, growing fast as humanoid and physical-AI labs race to close the real-world data gap that text and images cannot fill. It rewards spatial attention and careful, consistent annotation rather than a specific degree. Robotics companies also hire for this work directly, from teleoperating robots in a lab to test driving self-driving cars, and RobotTraining.jobs, a job board from the team behind this site, collects those openings from the employers' own job boards every day.