Mbodi AI wants to help robots adapt as fast as workers

This article is a part of Startup Spotlight, a series that features young, up-and-coming startups.

Image credit: Timmy Loen

While working together at Google, Xavier Chi and Sebastian Peralta saw a strange pattern in factories. Robots sat idle while people handled the constantly changing tasks, which pushed them to start Mbodi AI and rethink how robots learn.

😟 Problem

Most factories still rely on people for tasks that change frequently, also known as high-mix work. Industrial robots exist, but:

• Reprogramming takes weeks and specialist talent
• Small workflow tweaks stall automation projects
• Over 70% of factories report labor shortages even as labor remains nearly half of global GDP

High-mix manufacturing and logistics remain largely manual, despite heavy investment in hardware.

💡 Solution

Mbodi AI turns industrial robots into systems that workers can teach directly. Key features include:

• Operators give natural language commands such as “place three bottles in each tray” or demo a task once
• A network of AI agents interprets intent, plans safe motion, and executes on the robot within minutes
• Vision, reasoning, and motion control run together through a cloud-to-edge platform across different robot brands
• New skills sync across the fleet so one training session updates every connected robot

Image credit: Timmy Loen

📊 Market size

There are more than 5 million industrial robots in use and around 500,000 new installations each year.

If 10% of these robots handle high-mix work at roughly US$20,000 in annual software fees per unit, Mbodi AI targets about US$10 billion in recurring revenue. Each new deployment wave expands that base.

🤝 Team

• Xavier Chi co-founder. Former tech lead for Google Public DNS, which serves a core share of global internet traffic. Brings expertise in reliability-critical distributed systems and embodied AI.
• Sebastian Peralta co-founder. Physicist, roboticist, and deep learning researcher. Triple major in electrical engineering, computer science, and physics from UPenn with work at GRASP Robotics Lab and engineering experience on Google Public DNS.

🚀 Traction

• Commercialization partnership with ABB Robotics after winning its global AI startup challenge
• Paid proof of concept with a Fortune 100 consumer goods company for high-mix packaging tasks traditionally handled by humans
• Transition from pilots into production deployments
• Participation in Y Combinator Spring 2025 batch

🏆 Competition

💰 Financials