Industrial robots have been part of factories for decades, but they still carry one big problem. They are powerful, precise, and tireless, yet they can be difficult to teach when the work changes. A small product update, a new part shape, a different workflow, or a fresh warehouse layout can turn into weeks of reprogramming, specialist support, and production delays.
That is the kind of problem Sebastian Peralta is taking on with Mbodi AI. Instead of treating robots like machines that need rigid code for every task, Mbodi AI is working on a more flexible way to teach them. The company is building technology that lets industrial robots learn from natural language and quick demonstrations, making robot training feel closer to showing a person how to do a job than writing a long technical script.
For manufacturers, that idea matters. Factories and warehouses are under pressure to move faster, handle more product variety, and reduce downtime. The old model of automation often works well for repeated tasks, but it struggles when work becomes more dynamic. Sebastian Peralta and Mbodi AI are trying to close that gap by bringing modern AI into the physical world.
Who is Sebastian Peralta
Sebastian Peralta is a founder of Mbodi AI, a robotics and artificial intelligence startup based in New York. His background sits at the intersection of physics, robotics, deep learning, and large scale software systems. That combination is important because building smarter industrial robots is not just a software challenge. It also requires a deep understanding of how machines move, how sensors read the world, how AI models reason, and how systems perform under real pressure.
Before building Mbodi AI, Peralta worked in technical environments where reliability mattered. His experience includes engineering work connected to Google Public DNS, one of the major internet infrastructure systems used around the world. He also studied at the University of Pennsylvania, where his academic path touched electrical engineering, computer science, and physics. His connection to robotics research, including the GRASP Lab ecosystem, gives his work a strong foundation in both theory and real machine behavior.
That mix makes his founder story more interesting than a simple startup profile. Peralta is not only building an AI product around a trending market. He is working on a hard, physical problem where code has to meet metal, motion, sensors, safety, and factory reality.
What problem Mbodi AI is trying to solve
Most people outside manufacturing imagine industrial robots as highly advanced machines that can do almost anything. In reality, many factory robots are excellent at narrow, repeated tasks but far less flexible when the process changes.
A robot might be able to pick up the same part from the same place thousands of times a day. But when the object changes, the angle shifts, the workspace is rearranged, or the task sequence needs to be updated, the system may need specialist reprogramming. That can mean downtime, engineering cost, and delays.
This is especially painful in modern manufacturing because production is becoming more varied. Companies are dealing with smaller batch sizes, faster product cycles, changing customer demand, and more pressure to automate without locking themselves into one rigid workflow.
Mbodi AI is focused on that pain point. The company wants to make it easier for industrial teams to teach robots new tasks without needing deep robotics expertise every time. Instead of relying only on traditional programming, Mbodi AI is building an embodied intelligence layer that can understand instructions, learn from demonstrations, and turn that learning into robot actions.
In simple terms, the goal is to help robots become more adaptable.
Why industrial robots need a simpler way to learn
Factories do not run in perfect conditions forever. A process that works today may need to change next month. A packaging line might add a new product size. A warehouse might change its picking process. A manufacturer might introduce new materials or need to move robots between tasks.
Traditional automation can struggle in these settings because the robot does not automatically understand the new goal. It only knows what it has been programmed to do. That means teams often depend on robotics engineers, system integrators, or external consultants to adjust the setup.
For large companies, that can slow down decision making. For smaller manufacturers, it can make advanced automation feel out of reach.
This is where Sebastian Peralta’s work with Mbodi AI becomes useful. If factory teams can explain a task in plain language or give a quick demonstration, the learning process becomes more natural. A line operator does not need to become a robotics programmer to help the robot understand what needs to happen.
That shift could make industrial automation more practical for companies that need flexibility, not just repetition.
How Mbodi AI uses natural language and quick demos
The core idea behind Mbodi AI is easy to understand. A person should be able to teach a robot by describing the task and showing the action. The system then helps translate that instruction into reliable robotic behavior.
For example, instead of writing complex code, an operator might describe what the robot needs to pick, where it needs to place it, and how the task should be performed. A quick demonstration can provide extra context, such as the movement path, object handling, or desired end state.
Behind that simple user experience is a much harder technical challenge. The robot needs to connect several layers of intelligence. It has to understand language, interpret the physical scene, identify objects, plan movement, control the robot arm, respond to changes, and complete the task safely.
That is why Mbodi AI is not just about adding a chatbot to a robot. Industrial robotics needs AI that can act in the real world. A language model can understand a sentence, but a robot also needs perception, timing, motion control, and reliability. The system has to convert human intent into physical action.
This is where embodied AI becomes important.
How Sebastian Peralta connects robotics research with real factory needs
Embodied AI means artificial intelligence that can operate through a physical body or machine. It is different from AI that only writes text, summarizes data, or answers questions on a screen. In robotics, embodied AI has to understand the environment and take useful action inside it.
For Sebastian Peralta, this is the real challenge. Teaching a robot is not only about giving it instructions. The robot must know what those instructions mean in a changing physical space. It must handle uncertainty, adapt to variation, and still perform safely in a production environment.
That is why his background matters. A founder working in this space needs to understand AI models, but also the limits of physical systems. A robot cannot simply guess its way through a task. If it is working near people, expensive equipment, or delicate materials, its actions must be dependable.
Mbodi AI’s approach appears to focus on making robots easier to teach while still keeping the demands of industrial work in view. The company is not only aiming for impressive demos. It is building toward production use, where reliability is the difference between a clever idea and a real business tool.
Why Mbodi AI matters for manufacturing and warehouse automation
The promise of Mbodi AI is not just that robots could learn faster. It is that automation could become easier to scale across different jobs, machines, and facilities.
In many factories, one robotic solution is built for one specific task. If the company wants to automate another process, the work often starts over. Engineers analyze the task, create a new setup, program the robot, test it, and adjust it until it works. That process can be valuable, but it is slow.
If robots can be taught through natural language and demonstrations, manufacturers could shorten the path from idea to deployment. Teams could experiment faster. They could update tasks with less downtime. They could move robots into new roles without treating every change like a new engineering project.
This could be especially useful in areas such as:
- Pick and place operations
- Packaging and sorting
- Machine tending
- Warehouse handling
- Assembly support
- Repetitive material movement
- Quality related workflows
A more flexible robot learning system could also help companies build robotic skill libraries. Once a robot learns a useful action, that skill may be reusable in other workflows or across a fleet. Over time, this turns individual task training into a shared automation asset.
That idea is powerful because it makes robot learning less isolated. Instead of each deployment standing alone, knowledge can become part of a broader system.
The ABB connection and what it says about Mbodi AI’s potential
One reason Mbodi AI has attracted attention is its connection with ABB Robotics. Mbodi was named one of the winners of the 2024 ABB Robotics AI Startup Challenge, alongside T-Robotics. The challenge focused on AI solutions that could help industrial robots understand, learn, and adapt in more complex manufacturing environments.
That recognition matters because ABB is one of the major names in global robotics and automation. A startup working on industrial robots needs more than a strong pitch. It needs relevance to real factory problems, and recognition from a major robotics player gives Mbodi AI extra credibility.
The ABB connection also shows how timely this market has become. Robotics companies are looking for ways to make automation easier to deploy and more adaptable. Generative AI, natural language programming, and agent based robotics are becoming more important because manufacturers want robots that are not locked into narrow scripts.
For Sebastian Peralta and the Mbodi AI team, this creates a strong opening. The company is building at a moment when the industrial world is actively searching for more flexible automation tools.
What makes Sebastian Peralta’s founder story stand out
Many AI startups focus on software because software is easier to launch, test, and scale. Robotics is different. It is slower, harder, and less forgiving. A product has to work not only in a browser or dashboard, but also in a noisy, unpredictable physical environment.
That is what makes Sebastian Peralta’s path stand out. He is building in a space where technical depth really matters. Industrial robots need intelligence, but they also need control, safety, repeatability, and trust.
His work with Mbodi AI reflects a bigger shift in robotics. The goal is not only to make robots stronger or faster. The goal is to make them easier for people to work with. If a factory worker can teach a robot using everyday language and quick examples, the robot becomes less like a sealed machine and more like a flexible teammate.
That does not mean robots will suddenly become human. It means the interface between humans and machines can become simpler. The machine can stay precise and reliable, while the teaching process becomes more natural.
This is a practical version of AI progress. It is not about replacing every worker overnight or making unrealistic promises. It is about giving industrial teams better tools to handle repetitive, changing, and physically demanding work.
The bigger vision behind Mbodi AI
The long term vision for Mbodi AI appears to be a world where physical work can be automated with much less friction. Instead of every robot task requiring heavy manual programming, companies could build systems that learn, adapt, and improve over time.
In that kind of environment, a robot would not be limited to one fixed motion. It could develop a library of useful skills. A manufacturer could teach a task once, refine it through use, and apply that knowledge elsewhere. A warehouse could adapt robots to new products or workflows more quickly. A system integrator could deploy solutions faster because the robot already has a more intelligent learning layer.
This is where the idea of AI agents for the physical world becomes important. In software, AI agents can help complete tasks by understanding goals, choosing steps, and using tools. In robotics, that same concept becomes more complex because the tool is a physical machine. The agent must reason about real objects, space, motion, and risk.
Mbodi AI is working in that difficult middle ground between digital intelligence and physical execution. If the company can make that bridge reliable, it could change how industrial automation is bought, deployed, and improved.
The challenges of bringing embodied AI into industrial robotics
The opportunity is large, but the challenges are just as real. Industrial robotics is not a space where vague performance is good enough. Robots must work consistently, safely, and predictably.
One challenge is reliability. A robot that performs a task correctly nine out of ten times may still be unacceptable in a production environment. Factories need systems that can handle long shifts, repeated tasks, and edge cases without constant human rescue.
Another challenge is variation. Real factory environments include different lighting, dust, reflections, object positions, packaging changes, and unexpected interruptions. A system that works in a clean demo space still has to prove itself under daily operating conditions.
There is also the issue of trust. Operators and managers need to understand what the robot is doing and why. If a robot is learning from language and demonstrations, the system must make that learning transparent enough for teams to feel confident using it.
Cost is another factor. Even if the technology works, companies will ask whether it saves time, reduces downtime, improves output, or solves a labor problem that traditional automation could not handle as easily.
These challenges do not weaken the Mbodi AI story. They make it more grounded. The company is working on a hard problem, and the value will come from proving that AI taught robots can perform in real industrial settings.
Why Sebastian Peralta is becoming a name to watch in AI robotics
Sebastian Peralta is worth watching because his work sits inside one of the most important shifts in automation. Industrial companies do not only need robots that can repeat tasks. They need robots that can adapt when the work changes.
Mbodi AI is building toward that future by focusing on natural language, quick demonstrations, robot skill learning, and embodied intelligence. The company’s work speaks directly to a major pain point in manufacturing and warehouse automation, which is the difficulty of teaching robots new tasks without slow, expensive reprogramming.
Peralta’s background in robotics, deep learning, physics, and infrastructure gives him a strong base for this kind of company. His success so far is not just about launching another AI startup. It is about applying AI to one of the most difficult and useful areas of the real world.
If Mbodi AI continues to make progress, its impact could reach beyond one product or one robot. It could help reshape how factories think about automation itself. Instead of asking whether a task is worth programming from scratch, teams may start asking whether a robot can simply be taught.
