Robots can be programmed to do a variety of tasks, like packing boxes and even performing surgery. But each individual movement or task requires its own specific training process, which makes it hard for robots to adapt in real-world scenarios.
Mbodi wants to make training robots easier and quicker with the help of AI agents. The company will be showcasing this tech as one of the Top 20 Startup Battlefield finalists at TechCrunch Disrupt 2025.
New York-based Mbodi built a cloud-to-edge system, a hybrid computing system using both cloud and local compute, that is designed to integrate into existing robotic tech stacks. The software relies on a multitude of AI agents that communicate with each other to gather the needed information to help a robot learn a task faster.
Once deployed, Mbodi will collect data and learn from its real-world use cases.
Xavier Chi, co-founder of Mbodi, told TechCrunch that users prompt the software using natural language, and Mbodi breaks down the request into smaller subtasks. Mbodi’s cluster of agents essentially divides and conquers the task to gather the needed information to train the robot on the prompt quickly.
“The tricky thing with the physical world, it’s infinite possibility,” Chi said. “Every time you can invent something completely new, you haven’t had any data, that is a problem in the physical world. We always need to have a system where you can orchestrate different models or have anyone correct a robot and tell it to do certain things certain ways.”
Chi said he and co-founder Sebastian Peralta got the idea for the company while working as engineers at Google. While they weren’t working on robotics, they both came to the realization that the advancements in AI were heading to the physical world and despite a rise in physical AI, there still wasn’t a great way to quickly train robots.
Many companies, like Skild AI and FieldAI, are looking to help make training robots faster by building large world AI models with enough real-world data to make it easier for them to adapt to new environments. Chi said that philosophy just doesn’t work with how much the world constantly changes.
Mbodi launched in 2024 with a focus on picking and packaging. The company won an ABB Robotics AI startup competition last year, which landed them a partnership with the Swiss robotics organization that was acquired by SoftBank for $5.4 billion in October.
Now the company is working with a Fortune 100 company in the consumer and product goods space on a proof of concept.
“For the CPG customer, they have a lot of people, they pack different products of their brand into a tray or a shelf thing, the problem is it changes every day,” Chi said. “Because of that, it is impossible to put robots there. To reprogram these robots, it’s just not possible, there is still a lot of humans doing that work.”
Mbodi hopes to start deploying its software more in 2026.
“We want to build something that works, that can actually be deployed,” Chi said. “We aren’t a research lab; we don’t want to be a research lab in that regard. We want to put something in production that works reliably.”
If you want to hear from Mbodi firsthand, and see dozens of additional pitches, attend valuable workshops, and make the connections that drive business results, head here to learn more about this year’s Disrupt, held October 27 to 29 in San Francisco.
This piece has been updated to better reflect Chi’s title.