Mbodi will show how it can train a robot using AI agents at TechCrunch Disrupt 2025

The world of robotics is evolving rapidly, with machines being programmed to handle diverse tasks from packing boxes to performing intricate surgeries. However, training robots for each specific movement presents a significant challenge, particularly in adapting to the complexities of real-world environments. Enter Mbodi, a pioneering company aiming to streamline robot training through innovative AI agents. As one of the Top 20 finalists in the Startup Battlefield competition at TechCrunch Disrupt 2025, Mbodi is set to showcase its groundbreaking technology. Based in New York, the company has developed a hybrid cloud-to-edge computing system designed to enhance existing robotic technologies. This system utilizes a network of AI agents that communicate effectively to gather essential information, enabling robots to learn tasks more efficiently. Once deployed, Mbodi's technology gathers data from actual use cases, allowing for continual improvement. Co-founder and CEO Xavier Chi explained how users can interact with the software using natural language. The system intelligently breaks down requests into manageable subtasks, allowing its cluster of agents to tackle training challenges swiftly. Chi emphasized the complexities of the physical world, stating, "Every time you can invent something completely new, you haven’t had any data, that is a problem in the physical world." The need for a versatile system that can manage various models and accommodate corrections is critical in this space. The inspiration for Mbodi emerged while Chi and co-founder Sebastian Peralta were engineers at Google. Although they weren't directly involved in robotics, they recognized the potential of AI advancements in physical applications. While many companies are attempting to create large AI models to enhance robot training, Chi believes that this approach struggles to keep pace with the ever-changing real world. Since its launch in 2024, Mbodi has focused on optimizing picking and packaging processes. The company gained traction after winning an ABB Robotics AI startup competition, securing a partnership with a Swiss robotics organization recently acquired by SoftBank for $5.4 billion. Currently, Mbodi is collaborating with a Fortune 100 firm in the consumer goods sector to develop proof of concept. According to Chi, the consumer goods industry faces unique challenges, as product configurations change daily, making it difficult to rely on robots without reprogramming. "With constant changes, it’s impossible to automate these processes completely," he remarked, highlighting the ongoing need for human involvement. Looking ahead, Mbodi aims to deploy its software more extensively in 2026. Chi stated, "We want to build something that works, that can actually be deployed. We aren’t a research lab; we want to create reliable production solutions." For those interested in learning more about Mbodi's innovative approach and witnessing other pitches, workshops, and networking opportunities, the TechCrunch Disrupt event will take place from October 27 to 29 in San Francisco.