After actuator failures disrupted robotics deployments in Asia, Mahadewa began tracing the problem through components, materials, and manufacturing systems, experience that now informs the industrial company he is building in the United States.
Bihan Mahadewa is building a venture-backed industrial company focused on shortening supply-chain constraints that slow hardware iteration in the United States. The work grew from problems he encountered firsthand while trying to deploy LimX Dynamics’ TRON 1 robots across Asia, including in Indonesia and Malaysia. The robots were intended to help monitor and secure large environments that can be difficult or time-consuming for people to cover, such as industrial facilities and data centers. Actuator issues were serious enough to prevent successful deployments, pushing Mahadewa to look beyond the finished machine and investigate where the failures originated.
“That experience changed how I looked at robotics,” Mahadewa says. “The failure showed up at the robot level, but the cause could sit several layers deeper. If you only look at the finished machine, you can miss the industrial problem underneath it.”
When an actuator failed, the team had to wait for a replacement to be shipped from the original factory. Mahadewa says supplier responses were slow, and shipping delays extended troubleshooting and held up deployment. The team did not have the dimensions and technical files needed to build the replacement itself. The experience showed him how dependent a deployment could become on a single supplier and how difficult it was to find a faster sourcing alternative.
Mahadewa brought prior experience building and scaling physical products to that problem. In an earlier hardware company, he served as founder and CEO, led hardware development, built the engineering team, and helped scale the product to more than 30,000 units shipped. That experience gave him responsibility not only for product design, but also for what happened when a design had to move through suppliers and manufacturing systems at commercial scale.
His decision to build a team in China added another layer to that experience. Working closer to specialized suppliers and manufacturers gave Mahadewa a direct view of how concentrated production expertise could affect development speed. He personally worked on actuator designs and explored new configurations for robotics applications, building the designs from scratch. Direct supplier relationships also made it possible to source specialized parts, including custom magnets for brushless DC motors, more quickly.
Those magnets were not interchangeable components. Their geometry and magnetic properties could affect a motor’s torque characteristics and determine how the motor fit within a compact actuator. Working directly with suppliers let Mahadewa address those component-level decisions while the broader design was still evolving, reinforcing his view that manufacturing knowledge belongs inside product development.
The robotics deployment brought those lessons into sharper focus. Mahadewa began looking more closely at the physical systems behind robotic performance because the actuator problems could not be understood by examining software or the finished robot alone. Motors depend on magnets, while magnets depend on materials that must be separated, refined, converted, and manufactured consistently enough for demanding applications. Each layer introduced another place where a product could fail before deployment.
“Robotics gets discussed as if intelligence is the only hard part,” Mahadewa says. “But a very capable system still has to move the way you expect, survive repeated use, and be manufactured at a cost that makes deployment realistic.”
Following those dependencies led Mahadewa toward rare-earth supply chains and the industrial steps between raw materials and finished components. His interest is less in mineral access by itself than in the processing, production knowledge, and quality control required to turn materials into parts that can perform reliably. Separation, refining, metals, alloys, and magnet production each depend on specialized equipment and teams capable of keeping output within specification.
As founder and CEO, Mahadewa had already seen what happened when engineering decisions met production constraints at scale. The robotics failures gave him a more specific reason to follow those constraints farther upstream and ask which parts of the industrial system were limiting the speed at which engineers could test and improve physical products.
“Industrial capability is not something you buy in one transaction,” Mahadewa says. “You can purchase equipment, but the knowledge of how to run the process well is accumulated through production, mistakes, and repeated improvement.”
Mahadewa also came away from the robotics work convinced that real deployment reveals problems that demonstrations do not. Reliability, maintenance, integration, and cost become harder to ignore when a machine has to perform repeatedly outside a controlled setting. The failed actuator and the wait for a replacement illustrated how a component problem could quickly become a deployment problem when the team lacked the information and local capability to resolve it independently.
“Deployment is where the real bottlenecks show themselves,” Mahadewa says. “A demonstration can tell you that something is possible. It cannot tell you everything about what will break, what will wear out, or what becomes too expensive when you try to operate it every day.”
That experience now informs the industrial company Mahadewa is building. His focus is on reducing the supply-chain constraints that increase the time between identifying a hardware problem and testing the next version. Long component lead times, inconsistent production, or suppliers unable to meet a specification can reduce how many times an engineering team is able to build, test, and improve a product.
For Mahadewa, that makes iteration speed an engineering outcome as much as an operational one. At his earlier company, he led hardware development as the product scaled beyond 30,000 shipped units. The robotics deployments showed him how quickly progress can stop when one component or manufacturing dependency fails. His current work brings those experiences together around a narrower question: how can engineers get from a failed physical version to the next viable one faster?
“The best hardware rarely comes from getting everything right the first time,” Mahadewa says. “It comes from being able to learn quickly enough that each version is better than the one before it.”