Mbodi is set to demonstrate its AI-driven robot training system at TechCrunch Disrupt 2025. The system uses natural language prompts to enable users to teach robots new skills, leveraging a cluster of AI agents to simplify the training process. Mbodi's cloud-to-edge platform combines generative AI, agent orchestration, and symbolic reasoning to translate high-level instructions into precise and adaptive robot actions.
Mbodi's universal agent framework is compatible with various robotic hardware, coordinating perception, reasoning, planning, and control in real-time. When one robot learns a task, the knowledge can be shared across the entire fleet. This technology addresses the challenge of labour shortages in manufacturing, where traditional robot programming can be time-consuming and require specialised engineers. Mbodi aims to make robot training as easy as teaching a human, reducing the need for code and lengthy reprogramming.
The company's co-founders, formerly of Google, envision a future where robots are easily adaptable and accessible, turning them into consumer commodities. Mbodi is currently rolling out production deployments with partners like ABB Robotics, automating complex tasks in real-world facilities. Mbodi was also the winner of the 2024 ABB Robotics AI Startup Challenge and a 2025 RBR50 Robotics Innovation Award.
