Key Points
- Mbodi founded by former Google engineers Xavier Chi and Sebastian Peralta.
- Hybrid cloud‑to‑edge platform uses multiple AI agents to train robots via natural‑language prompts.
- System breaks user requests into subtasks, enabling faster robot adaptation for picking and packaging.
- Won ABB Robotics AI startup competition and partnered with a Swiss robotics firm later acquired by SoftBank.
- Currently piloting technology with a Fortune 100 consumer‑goods company.
- Will showcase the solution at TechCrunch Disrupt 2025 in San Francisco (October 27‑29).
- Plans for broader deployment begin in 2026.
TechCrunch Disrupt 2025 no anniversary
Mbodi’s Vision for Smarter Robot Training
Mbodi was founded by Xavier Chi and Sebastian Peralta, engineers who previously worked at Google. While at Google they recognized that advances in artificial intelligence were moving toward physical applications, yet there was no efficient way to train robots for new, dynamic tasks. Their answer is a hybrid cloud‑to‑edge platform that integrates with existing robotic stacks and relies on a network of AI agents that communicate to gather the information needed for rapid robot training.
How the Platform Works
The system accepts natural‑language prompts from users. Once a request is received, Mbodi’s software decomposes the request into smaller subtasks. Each AI agent tackles a portion of the problem, collecting data, generating instructions, or refining motions. By dividing and conquering, the agents accelerate the learning loop, allowing robots to adapt to tasks such as picking, packaging, or even more complex operations without extensive reprogramming.
Key Milestones and Partnerships
Since its launch in 2024, Mbodi has focused on the picking and packaging segment. The startup earned recognition by winning the ABB Robotics AI startup competition, which secured a partnership with a Swiss robotics organization that was later acquired by SoftBank. Building on that momentum, Mbodi is currently conducting a proof‑of‑concept deployment with a Fortune 100 consumer and product‑goods company, addressing the challenge of constantly changing packaging configurations that make traditional robot reprogramming impractical.
TechCrunch Disrupt Showcase
Mbodi will present its technology as a finalist in the Top 20 Startup Battlefield at TechCrunch Disrupt 2025, held in San Francisco from October 27‑29. The event will give the company a stage to demonstrate how its AI‑agent framework can reduce the time and effort required to train robots for new tasks, positioning Mbodi as a practical alternative to research‑lab‑only solutions.
Future Outlook
The founders emphasize that Mbodi is not a research lab; they aim to deliver a production‑ready system that works reliably at scale. While the company plans broader deployments in 2026, its immediate focus is on refining the platform through real‑world use cases and expanding partnerships across industries that demand flexible, adaptable robotic solutions.
Source: techcrunch.com