Rudy Cohen is the CEO and Co-Founder of Inbolt, a startup bringing vision and AI to robotics in manufacturing. He began his academic journey as a researcher in applied mathematics at École Normale Supérieure Paris-Saclay, before completing a Master’s in entrepreneurship from École Polytechnique and HEC Paris. Inbolt is building the real-time control layer that turns digital twins into live robot execution. Since its founding in 2019, the company has raised €20 million and now operates in over 100 factories across Europe, the US, and Japan.
Physical AI's consensus says progress is gated by data. More of it, better curated, better simulated. That's true in the lab. In the factory, it's the wrong constraint. After deploying vision-guided control across 100+ factories and 40 million-plus robot cycles, Inbolt CEO Rudy Cohen argues the real bottleneck is the loop between perception and motion, and the integration tax surrounding it.
This talk reframes where Physical AI actually pays off. Not in bigger models, but in closing the loop at servo frequency and collapsing the months of fixtures, wiring, and retuning that make every deployment a six-figure project. Using production data from lines at Stellantis, Toyota, and Ford, Cohen makes the case that Physical AI is already commercially proven, just not where the headlines are looking.
Key takeaways:
