OriginFlow builds embodied-AI data infrastructure based on human manipulation signals. The company uses wearable sensing and human-motion capture to turn real human skill into training data for robots, supporting physical-AI systems that need scalable demonstrations for dexterous work.
About
OriginFlow builds data-capture systems for training embodied-AI models from human manipulation. Its NeuroScale approach combines surface electromyography, inertial measurements and first-person vision to reconstruct motion, pose, force, contact and operator intent.
The Origin Kit packages a neuromuscular wristband, first-person capture device and control unit for synchronised recording during real work. OriginFlow turns those recordings into training assets; robot hardware, policy training and transfer to a target embodiment remain separate stages.
OriginFlow is relevant because dexterous robot behavior is constrained by data quality, not only hardware. The company is building a data layer for fine manipulation, home-service skills and physical interaction datasets that robot builders can use to train…