Unveiling Mbodi’s Groundbreaking AI Training Methods for Robots at TechCrunch Disrupt 2025

Redefining Robot Training with AI: Mbodi’s Innovative Approach

Robots have the potential to revolutionize various industries by performing tasks ranging from packing boxes to conducting complex surgeries. However, training robots for each specific movement or task can be a time-consuming process, limiting their adaptability in real-world scenarios.

Mbodi, a New York-based company, aims to streamline and accelerate the robot training process using AI agents. This groundbreaking technology will be showcased as one of the Top 20 Startup Battlefield finalists at TechCrunch Disrupt 2025.

Mbodi has developed a cutting-edge cloud-to-edge system that combines cloud and local computing to seamlessly integrate into existing robotic technology stacks. The system relies on a network of AI agents that collaborate to gather essential information, enabling robots to learn tasks more efficiently.

Once deployed, Mbodi’s system continuously collects data and learns from real-world usage scenarios, further enhancing its capabilities.

Xavier Chi, the co-founder of Mbodi, explained that users interact with the software using natural language, which is then parsed into smaller subtasks by Mbodi. The system’s cluster of AI agents collaborates to gather the necessary information swiftly, facilitating quick training of robots based on user prompts.

Chi emphasized the challenge posed by the infinite possibilities in 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.” Mbodi’s approach of orchestrating different models through AI agents addresses this challenge effectively.

The inspiration for Mbodi stemmed from Chi and co-founder Sebastian Peralta’s experience as engineers at Google. Despite the advancements in AI, they recognized the need for a more efficient method to train robots rapidly, leading to the inception of Mbodi.

Empowering Robots for Dynamic Environments

Several companies, including Skild AI and FieldAI, are also working on enhancing robot training by leveraging large world AI models with extensive real-world data. However, Chi pointed out that this approach may not adequately address the constantly changing nature of the world.

Mbodi made its debut in 2024 with a focus on picking and packaging tasks. The company’s innovative solutions garnered recognition when they won an ABB Robotics AI startup competition, leading to a partnership with a Swiss robotics organization acquired by SoftBank for $5.4 billion in October.

Currently collaborating with a Fortune 100 company in the consumer and product goods sector on a proof of concept, Mbodi aims to revolutionize tasks that involve daily variations, making them challenging for traditional robotic solutions.

Looking ahead, Mbodi plans to expand the deployment of its software in 2026, with a strong focus on practical, reliable solutions. Chi emphasized the company’s commitment to delivering a deployable product rather than remaining a research entity.

If you’re interested in learning more about Mbodi’s groundbreaking technology and exploring a range of innovative pitches, workshops, and networking opportunities, don’t miss this year’s Disrupt event, scheduled for October 27 to 29 in San Francisco.

Don’t miss the opportunity to witness firsthand the future of robotics and AI at TechCrunch Disrupt 2025!