He spent his twenties optimizing rocket structures. Now he is teaching machines to run reality in software - before anyone builds the real thing.
Walk into Pasteur Labs and you will not find a typical AI startup chasing the latest chatbot. Alexander Lavin runs something stranger and more ambitious from New York: a deep-tech lab whose product is, in his words, "the IDE for reality." The pitch is simple enough to fit on a napkin and audacious enough to take a decade. Most engineering still runs on slow, expensive, real-world testing - wind tunnels, prototype reactors, crash sleds. Lavin wants to flip the ratio. "What if we could flip that around," he asks, "so 90% of your expensive, painfully slow, real-world testing is tackled in software?"
To get there, Pasteur Labs fuses physics simulation with machine learning into what the field now calls simulation intelligence - in-silico playgrounds where human-machine teams can model multi-physics, multi-scale phenomena and ask cause-and-effect questions that a pure neural network would only guess at. The targets are not toys. They are nuclear fusion, energy security, climate, and the industrial R&D that keeps the lights on.
Alongside the for-profit lab sits its sibling, the Institute for Simulation Intelligence (ISI), a 501c3 nonprofit. The split is deliberate: Pasteur builds and ships, ISI supports the broader scientific ecosystem in the non-commercial ways a company never could. Both point at the same north star Lavin likes to call "Nobel-Turing technologies" - tools that advance science and society at the level of a Nobel discovery and a Turing breakthrough at once.
What makes Lavin worth a second look is not a single triumph. It is the pattern. He is a founder who keeps leaving fields he is already winning. Spacecraft engineering. Theoretical-neuroscience AI. AGI research. Probabilistic medicine. Earth-systems modeling. Each move looks like a non-sequitur until you notice the through-line: he is hunting for the place where computation, physics, and intelligence collide hardest.
The phrase he repeats is almost a creed. "Engineers advance technology and technology advances civilization." It is the kind of line that could sound like a poster in a co-working space. From Lavin it reads more like a job description he has been filling out, role by role, for fifteen years.
“What if 90% of your expensive, painfully slow, real-world testing is tackled in software?” — Alexander Lavin, on the premise of Pasteur Labs
The lunar rover came first. As a master's student at Carnegie Mellon, Lavin led student teams designing rovers for the Google Lunar X-Prize, then wrote a thesis on spacecraft structure optimization in partnership with Blue Origin. He interned at NASA Ames. He led capsule design at Hyperloop Transportation Technologies. For a stretch, the resume reads like a sci-fi character sheet.
His verdict on the place that shaped him is unsentimental and exact: "Carnegie Mellon is the most intellectually stimulating environment I've been a part of." He credits mentors Red Whittaker - the legendary field-robotics pioneer - and Kenji Shimada. And he took a lesson from outsiders who won anyway, citing Nest's Matt Rogers as proof that you can succeed in a domain you did not start in.
"You can learn literally anything you want, given you work at it, so do what you're passionate about." It is the rationale for a career that refuses to stay in one lane - and the reason he could jump from rocket propulsion to probabilistic machine learning without flinching.
In December 2021, Lavin and two dozen collaborators - including complexity scientist David Krakauer and Hector Zenil - dropped a 100-plus-page document that read less like a paper and more like a blueprint. "Simulation Intelligence" argued that scientific computing, simulation, and AI were converging into a single discipline, and laid out nine "motifs" as its building blocks: from differentiable physics and probabilistic programming to causal reasoning and open-ended optimization.
The applications were not modest: inverse problems in synthetic biology, directing nuclear-energy experiments, predicting emergent behavior in entire economies. It became one of the reference texts for the AI-for-science movement, and the intellectual spine of everything Pasteur Labs builds.
Relative emphasis across the motifs Pasteur leans on most — illustrative, not a benchmark.
“Engineers advance technology and technology advances civilization.” — The line Lavin keeps coming back to
An outdoor enthusiast and dog owner whose down-time looks a lot like recovery from carrying five careers at once.
His GitHub handle nods to the cosmology thought experiment about a self-aware mind assembling itself out of chaos. His Twitter handle winks at Rick and Morty's dimension C-137.
The lab's name salutes Louis Pasteur and "use-inspired basic research" - science that is rigorous and useful at the same time. The whole company is a footnote to a famous diagram.
Sean Carroll for physics, Richard Dawkins for evolution, David Foster Wallace for prose, Liu Cixin for the far future. A taste that spans the equation and the novel.
Forbes called at 26. Colleagues describe him as humble and determined anyway - more interested in the next hard problem than the last headline.
Rockets, neuroscience, AGI, medicine, climate. Each pivot looked random; together they trace a single hunt for where computation meets the physical world.