Building the Frontier Lab for Industrial Engineering

Our Chief Scientist Johannes Brandstetter on why Emmi AI is joining Mistral AI and what this next chapter means for Engineering AI and industrial innovation.

‍

Emmi AI is joining Mistral, and together we are setting out to build the first frontier lab for industrial engineering. This is the moment we go all in.

Why this, why now

More than six years ago, when I started to use neural networks to learn solutions of partial differential equations, this was a niche pursuit. However, many knew already back then: PDEs are the language of physics, they describe how fluids flow, how structures bend, how heat moves, how the physical world actually behaves. PDEs are the cornerstone of science and engineering, and therefore of every industry that builds physical things. I never doubted for a second that AI would eventually transform this field. What I did not see coming was how fast it would happen.

Meeting Dennis Just and Miks Mikelsons was one of the best things happening in my professional life. They encouraged me to go big. They provided enterpreneurial skills I was duly missing. And last but not least, they helped me to bring ivory tower research into the spotlight of frontier labs. When we founded Emmi AI in December 2024, we set out to pioneer new ground. In a year and a half, our team has delivered things I am genuinely proud of. We unlocked industry-scale computational fluid dynamics, and multi-physics modeling of turbulent, large-scale systems with direct industrial relevance, the kind of problems that have resisted clean solutions for decades. We also pioneered fully data-driven AI approaches that generalize across engineering verticals with minimal structural intervention. And just as importantly, we have always been a "science first" and "no bullshit" company. That is what has set us apart.

At the core of Emmi AI is an exceptional team. To everyone at Emmi: what you have built over the last months is extraordinary, and none of what comes next would be possible without you.

What changes, and what doesn't

What changes is the scale of what we can attempt. Mistral brings a world-class science team, a fast-growing applied science organization, the compute we need, and enterprise customers who want engineering transformations. Everything of that at a scale that was unthinkable one or two years ago. The signals are unambiguous. The time to turn computer-aided design, computer-aided engineering, workflow automation, and digital twins upside down is now.

What does not change is the mission. We are still building foundational intelligence for engineering. We are still building digital twins that capture real physics. We are still building toward agentic engineering, systems that do not just simulate but reason, design, and iterate alongside engineers. The ambition is the same. The resources, opportunities, customers, and team behind it are now in a different league.

Last but not least: We are opening a Mistral office in Linz and start building from there. More to come!

To our customers, partners, and the field

To the engineers and companies who took a bet on us early, thank you. You will see what that bet has brought you very soon. To the research community we grew up in, at JKU and in the broader ML and scientific computing world, this is your win too. The ideas did not come from nowhere.

To everyone reading this who builds, designs, simulates, or engineers physical systems for a living: the next few years are going to be the most interesting your field has seen in a generation. We intend to be the lab that defines them.

Emmi AI is special. Joining Mistral is how we give it the future it deserves.

Let's build!

Johannes

‍