New York-based Mbodi is set to showcase its artificial intelligence (AI) agent-driven system, designed to accelerate robot training and adaptation, at TechCrunch Disrupt 2025. The company has been named a Top 20 Startup Battlefield finalist for the event, which will take place in San Francisco from October 27 to 29, 2025.
Mbodi has developed a cloud-to-edge hybrid computing system that integrates with existing robotic tech stacks. This software utilizes multiple AI agents that communicate to gather necessary information, facilitating faster task learning for robots. Xavier Chi, co-founder and CEO of Mbodi, stated that users interact with the software using natural language prompts, which the system then breaks down into smaller subtasks. According to Chi, this approach allows Mbodi's cluster of agents to "divide and conquer" the task, thereby expediting the robot's training.
The company was founded in 2024 by Chi and Sebastian Peralta, both former Google engineers, who identified a need for quicker robot training solutions amidst advancements in AI and its application to the physical world. Mbodi initially focused on applications in picking and packaging. The firm received an ABB Robotics AI startup competition award last year, leading to a partnership with the Swiss robotics organization, which SoftBank later acquired for $5.4 billion in October 2025.
Mbodi is currently engaged in a proof-of-concept project with a Fortune 100 company in the consumer and product goods sector. Chi explained that the dynamic nature of product configurations in this industry often makes traditional robot reprogramming impractical, leaving many tasks to human labor. He noted the challenge in the physical world where "every time you can invent something completely new, you haven't had any data." Mbodi aims for broader deployment of its software in 2026, with Chi emphasizing the company's objective to "build something that works, that can actually be deployed," differentiating it from a research-focused entity.
