About the role
You will provide high-fidelity training data necessary for Vision-Language-Action (VLA) models to bridge the gap between human reasoning and robotic physical execution. This is a fixed-term, 5-6 month engagement focused on large-scale Real2Sim data infrastructure.
Key responsibilities
- Perform repetitive pick-and-place and dexterous manipulation tasks according to a strict, scripted workflow (e.g., sorting items, folding laundry, or operating appliances)
- Execute all movements in a specific robotic style, characterised by slow, precise arcs, to ensure the data is compatible with robot kinematic constraints
- Wear and maintain a PICO 4 Ultra AR/VR headset and five motion trackers (placed on the waist, ankles, and forearms) throughout the collection session
- Follow mandatory gestural triggers, including clear Start and End hand gestures for every 1-minute task sequence
- Ensure hands and objects remain consistently within the designated camera frame to meet high-fidelity quality standards
About you
- Ability to stand and perform fine-motor tasks
- Reliability for a full 5-6 month term engagement to support consistent model training
- Ability to sustain focus on scripted task execution without reverting to natural human movement speed
