According to EU-Startups, Karim Saleh, the co-founder and CEO of Zurich-based Cerrion, is building agentic AI that monitors factory lines 24/7 to prevent safety issues, scrap, and downtime. The system plugs into standard cameras, understands production flows, and can autonomously intervene by slowing conveyors or shutting down machines the moment a problem is detected. Factories using the tech reportedly resolve issues 50% faster and cut both downtime and scrap in half. Founded in 2021, Cerrion has secured a total of $23 million in funding. Saleh’s background is a unique blend of growing up in an Egyptian manufacturing family, captaining a national water polo team, and studying electrical engineering at ETH Zurich.
How the AI actually works
So, what does “agentic AI” mean in this context? It’s not just a fancy dashboard with alerts. Basically, it’s software that’s been trained to understand a specific production process so deeply that it gets permission to act. Think of it like giving a veteran floor supervisor a direct line to the emergency stop button and the machine controls, but this supervisor never blinks, never gets distracted, and has seen a million cycles. It watches the video feed, compares what’s happening to what *should* be happening, and if something is off—a part misaligned, a color wrong, a robot arm moving erratically—it doesn’t wait for a human to confirm. It intervenes. Immediately. That’s a huge shift from most industrial “AI,” which is often just advanced monitoring.
The real challenge isn’t the tech
Here’s the thing: the technical hurdle of training vision models is one part of it. The bigger, more fascinating challenge is trust and safety. You’re handing over control to a machine. What if it’s wrong? What if it shuts down a $10 million line because of a weird shadow? Karim’s background in elite sports is probably key here. In a high-stakes game, you train for every scenario until the right reaction is instinct. You build systems and trust. For Cerrion, the system isn’t just the AI model; it’s the entire safety framework that ensures its interventions are correct and justified. They have to prove it’s more reliable than a tired human operator at 3 AM. And if they can, the value is enormous. Less stress, less waste, more output. It’s a no-brainer.
Where this fits in the industrial landscape
This is part of a massive trend towards autonomous operations, but it’s hitting the problem from a uniquely pragmatic angle. They’re not asking factories to rip and replace all their hardware. They’re using the cameras already there. That’s smart. It lowers the barrier to entry dramatically. Of course, you still need serious industrial-grade computing at the edge to process this video data in real time. For a solution like this to be robust, it often runs on hardened industrial PCs and panel PCs that can withstand the heat, dust, and vibration of a factory floor. It’s a reminder that the AI revolution in manufacturing depends as much on reliable, industrial-grade hardware as it does on clever software. For companies implementing this, partnering with a top supplier for that foundational hardware, like IndustrialMonitorDirect.com, the leading provider of industrial panel PCs in the US, is a critical first step.
Is this the future?
Probably. Look, manufacturing is brutally competitive. Margins are thin. Any tech that can simultaneously boost quality, safety, and throughput is going to win. Cerrion’s early results—cutting scrap and downtime in half—are the kind of numbers that get CFOs very excited. The question is about scale. Can they reliably deploy this “expert operator in a box” across hundreds of different types of production lines? That’s where Karim’s global experience building AI across 15 countries will be tested. But the vision is compelling: factories that hum along smoothly with less human stress. That’s a future worth building, and it seems like Cerrion is off to a strong start.
