InstAdapt develops embodied-world-model infrastructure for Physical AI, spanning robot data production, 3D scene understanding and generation, simulation learning, model training and policy optimisation.
About
InstAdapt is a ShanghaiTech University spinout building infrastructure intended to help robots acquire reusable experience across changing environments, tasks and embodiments. Its technical scope covers multimodal robot-data pipelines, embodied world models, 3D scene understanding and generation, simulation learning and robot-policy optimisation.
The company reports an end-to-end pipeline from data production and physical-asset generation through model training and robot deployment. It has validated the stack in dexterous manipulation and fine-task scenarios and presented its Physical-AI infrastructure work with NVIDIA at GTC 2026.