Look around the room. Almost everything in it was shaped by a machine, and those machines were most likely not running at their best when they made it. On any single part the inefficiency is too small to notice. Repeated across every part, every machine, and every factory running at this moment, it becomes one of the largest and quietest sources of waste there is.

An injection-molding line we optimized makes more than 40 million PET bottles a year. With "good enough" process parameters, it was putting half a gram more plastic than needed into each bottle. That's about 20 tons of plastic on a single line. The world makes over 300 billion PET bottles each year on lines like it. "Good enough" is not sustainable.

When fine-tuning the parameters of an industrial machine, a real expert lands excellent parameters in a handful of runs. Everyone else settles for good-enough parameters after dozens of trial-and-error runs.

Dr. Jonathan Spitz, our founder, started in mechanical engineering and moved into robotics, then the AI behind it. Getting reliable results from very little real-world data became the thread of that work, first at Inria, France's national institute for AI, where the #SmallData approach began, and then at the Bosch Center for Artificial Intelligence, where he turned to AI for manufacturing.

There the problem came into focus. The field was aiming its heaviest tools, deep learning, more data, more compute, at an industry that has almost none of the data they need. A factory cannot run ten thousand experiments to find one setting. Every run costs material, time, and a machine that should be making parts. Manufacturing didn't need a bigger model. It needed AI built for few, costly experiments. So he left to build it. GaussML, and the team that has grown around it, has been proving it on real machines ever since.

Since Optimyzer launched, we have delivered better parameters across a wide range of manufacturing processes, and value within weeks. We plan to deliver ten times more value for our customers.

Complexity keeps rising, and the expertise to manage it is scarce across the entire value chain, not just at the machine. Conventional AI struggles there: the deep-learning playbook depends on knowledge-rich data a factory never generates, and pilot after pilot fails to pay off. Our #SmallData approach ensures our solutions deliver value by efficiently building the needed knowledge. That is what will carry our approach across the factory floor into planning, maintenance, logistics and beyond.