Mbodi, a startup based in New York, is set to showcase its innovative approach to robot training at TechCrunch Disrupt 2025. The company has developed a cloud-to-edge system that integrates with existing robotic technology stacks, leveraging AI agents to streamline the learning process for robots.
Mbodi’s software interprets natural language prompts and breaks them down into manageable subtasks. The collaborative effort of the AI agents facilitates rapid training by efficiently acquiring the necessary data to instruct the robot. Co-founder Xavier Chi emphasized the challenge of infinite possibilities in the physical world and the need for a system like Mbodi to orchestrate diverse models for effective robot training.
Co-founders Chi and Sebastian Peralta were inspired by their time at Google, recognizing the impending integration of AI into physical domains. Despite the evolution of physical AI, they identified a gap in efficient robot training methods, leading them to create Mbodi.
Source: TechCrunch