Robots can be programmed to perform a variety of tasks, such as packing boxes or performing surgery. However, each individual movement or task requires a unique training process, making it difficult for robots to adapt to real-world scenarios.
Mbodi wants to make training robots easier and faster with the help of AI agents. The company plans to showcase the technology at TechCrunch Disrupt 2025 as one of the Top 20 Startup Battlefield finalists.
New York-based Mbodi has built a cloud-to-edge system, a hybrid computing system that uses both cloud and local computing, designed to integrate into existing robotics technology stacks. The software relies on a large number of AI agents that communicate with each other to collect the information needed to help the robot learn tasks faster.
After deployment, Mbodi collects data and learns from real-world use cases.
Xavier Chi, co-founder and CEO of Mbodi, told TechCrunch that users prompt the software using natural language, and Mbodi breaks down requests into smaller subtasks. Mbodi’s cluster of agents essentially divides and conquers the task of following prompts and gathering the information needed to quickly train the robot.
“The difficult thing about the physical world is that there are infinite possibilities,” Chi said. “Anytime you can invent something completely new, you don’t have any data at all. That’s the problem with the physical world. You always need a system where you can tweak different models, or someone can modify a robot and tell it to do certain things in certain ways.”
Chi and co-founder Sebastian Peralta said they got the idea for the company while working as engineers at Google. Neither of us worked in robotics, but we realized that advances in AI were moving toward the physical world, and that despite the rise of physical AI, there still wasn’t a good way to train robots quickly.
tech crunch event
san francisco
|
October 27-29, 2025
Many companies, such as Skild AI and FieldAI, aim to speed up the training of robots by building large-scale global AI models with enough real-world data to make it easier to adapt to new environments. Chi said philosophy doesn’t work when the world is constantly changing.
Mbodi was founded in 2024 with a focus on picking and packaging. Last year, the company won the ABB Robotics AI startup competition and secured a partnership with robotics organization Swizz, which was acquired by SoftBank for $5.4 billion in October.
The company is currently working on proofs of concept with Fortune 100 companies in the consumer goods and products space.
“For CPG customers, they have a lot of employees and they have trays and shelves filled with different products from their own brand. The problem is that it changes every day,” Chee said. “So it’s impossible to put robots there. It’s impossible to reprogram these robots. There are still a lot of humans doing that work.”
Mbodi hopes to start rolling out more software in 2026.
“We want to build something that works and can actually be deployed,” Chi says. “We’re not a lab. We don’t want to be a lab for that matter. We want to put things into production that we know will work.”
To hear directly from Mbodi, see more pitches, attend valuable workshops, and make connections that drive business results, visit here to learn more about this year’s Disrupt, taking place in San Francisco from October 27th to 29th.
