Stories of Activate Fellows and alumni turning scientific breakthroughs into hard-tech impact.
Even knowing exactly how his company’s software works, Nosa Edoimioya (Reforge Robotics, Cohort 2024) admits that seeing it work in real life is still “kind of magical.”
“First you see the robot move without our software, and then you see it move with our software—and it's two different things,” he says. “We haven’t physically changed anything. Something underneath is happening. It feels like a brain or something.”
Reforge’s first-of-a-kind advanced motion-control software eliminates these pitfalls. Once installed, it immediately increases the speed and precision of industrial robots, with no changes to the machine itself—kind of like magic.
This advance could accelerate cutting-edge applications for robotics—not just picking something up and moving it from here to there, which Edoimioya says is 80-90% of what robots currently do. Reforge’s first customers are making robots for biomanufacturing, lab automation, and cable handling, where precision is key.
Most of all, Edoimioya hopes better-performing, less costly robots will make automation far more feasible, ultimately enabling the United States to build back its manufacturing capacity.
Founded in 2024, Reforge is now a four-person company, backed by $1.7M in pre-seed funding, and serving seven robotics companies. Carving its own path as a software-first motion-control platform for robot manufacturers and integrators, Reforge is currently demonstrating use cases with its early customers and building out a library of datasets.
Edoimioya recently wrapped up his two years as a fellow in the Activate Berkeley community. We sat down with him to reflect on his founder journey, get excited about Reforge’s potential impact, and hear his advice for the next generation of innovators.
What makes Reforge different from existing solutions?
Here’s the main difference. The way you would do this before us is you would fine-tune the robot program in order to accomplish a specific task, and you might spend hours doing this. In fact, one of our early customers told us that they presumed that they would spend 5% of engineering time perpetually fine-tuning motion, speed, and things like that. And if you disconnect the robot or it loses power, you have to do it all again.
It’s been sort of accepted in the robotics community that there's just a lot of fine-tuning that goes into building different applications. Reforge turns motion and sensor data into models of how an individual robot actually behaves. Our Shaper product reduces residual vibration and settling time. Joint Tracker reduces dynamic toolpath-tracking error. KineCal corrects kinematic error to improve positional accuracy. You get all of this right out of the box and you eliminate all that fine-tuning time.
What’s the potential impact you envision for Reforge?
We’re really excited about the opportunity to rebuild a manufacturing industrial base in the U.S. through automation.
Right now, there are almost 400,000 factories in the U.S., and we think there could be double that.
But we currently don't have enough hands to do this—manufacturing in America is already understaffed. What better way to increase the number of hands than to have the people who are currently doing it and already have a lot of knowledge to be able to manage a team of robots and AI models that can now support them? I think that’s how we grow the pie.
Tell us about the videos you’ve been posting that show how your software works in the real world.
We've been developing these modules with our customers, and now we want to show off what they can do.
So this one that I posted recently shows our vibration control module, which addresses the speed side of robotics. So let’s say you’re working in automotive manufacturing, where a 5% improvement in cycle time is a really big deal when you're making millions of parts a year, you can use our software to make the robot move faster.
What we showed in this video was one of our customer's robots—a company called Standard Bots. You could see that as we were increasing the speed of the robot, there's a linear increase in vibration of the robot as well.
But when you apply our software, the vibration stays flat. So you can basically increase the speed and not have any increase in vibration. And so you get a two-times increase in speed with no cost of hardware. You just put a software on and it controls the robot for that. So, that's one big capability, and we have some more coming down the pike.
What gains in accuracy are you seeing, and what difference could that make?
The first thing I described was speed—being able to move the robot two times faster with no increase in errors or vibration.
Our second main focus is improving the accuracy of the robot, which we've been working on with a robot manufacturer called UFactory, and we have a demo that increases position accuracy of the robot by 10 times.
If you’re using a robot to insert an Ethernet cable into a server, for example, that could be the difference between missing the insertion or actually getting the insertion right. And those are the kinds of errors that make an automation system infeasible—if the robot misses the insertion at all, the operator has to come and fix it, and they may decide, "Okay, this is actually wasting my time and I'm just going to do it myself."
This kind of scenario represents a lot of the reasons why automation doesn't work. But if we can solve that gap and make sure it's going to insert that Ethernet cable where you want it to go 100% of the time, that's a big deal for customers.
How did your vision for impact change over the course of the fellowship?
Two years ago, we knew we wanted to make robots faster and more accurate and save money for manufacturers.
But through this process, and talking with my teammates and people in the industry, we realized that we have an opportunity to solve an even bigger problem, which is “sim to real.”
When you run a simulation of a robotic cell or system, there are always some differences between that simulation and the real world because there are all kinds of manufacturing errors that the simulation can’t see. This is a canonical problem in robotics that everyone deals with, and it costs a lot of time and money.
We want to take the small solution we have now, continue to build it up, and make it so robust that you can have perfect one-to-one mapping from the computer to the real world.
Accomplishing this would have all kinds of implications for the industry. One: reducing the time that it takes to deploy a robot. Two: in the context of AI, generating data in simulation that actually matches the real world would mean that you can train on much more data—and you don't have to go and collect it individually. Right now, there are a lot of people collecting data for robotics, and it's taking up a significant amount of money and time.
How was the Activate Fellowship different from what you expected?
I thought it was just going to be technology development. I remember in both my application and finalist presentation saying, "Okay, here are the technological milestones that we're going to address over the course of this time."
We did progress technologically during the fellowship. But what I actually learned the most was: how does what I want create value for other people? One of the core tenets of Activate and building companies is figuring out, “How does what I want to do help other people do what they want to do?”
So, I entered expecting two years of technical development. What I learned was to ask how the technology helps other people accomplish what they want to do.
What’s your advice for getting the most out of your two years at Activate?
The most impactful thing for me as an Activate Fellow has been getting to know myself—and this may be the most impactful part of being a founder in the first place.
I spent a lot of time talking with the Activate managing directors and fellowship managers on how I'm doing and what I'm thinking about. I’ve also invested a lot in professional development, including a storytelling coach, for example. There was a point when I think I was working with three coaches at the same time.
It takes a lot of personal reflection and help from others to discover who you are and who you want to be—but it is transformational. Once you do that, then you have a tremendous amount of clarity about where you're going.
What would you tell a kid who dreams of building a cutting-edge technology?
The advice I would give to baby Nosa or another kid is that it is true—that thing you're really interested in, you can make a life out of doing that. There are more avenues than ever to take your interest and curiosity and turn it into something that can create value for people. It's really about deciding that you're going to overcome your fear about whether you can do this or not, and just start doing it and giving it your all.
Who should reach out to you today?
Reforge wants to hear from robot manufacturers and precision integrators with a specific vibration, tracking, or positional-accuracy problem.
Bring us one representative trajectory or error dataset, and Reforge will identify the relevant product, required inputs, and a bounded evaluation plan.