Olfactory Intelligence // Osmo
Alex
Wiltschko
He gave computers the one sense they never had. A nose.
Founder & CEO, Osmo. Eighteen years chasing one question: why does a molecule smell the way it does?
He gave computers the one sense they never had. A nose.
Slice a fresh summer plum. Drop it into a device that sniffs its molecules, push the reading to the cloud, and in a different room a printer mixes the smell back into the air. Bite-into-it accurate. Osmo calls it scent teleportation, and in 2024 it became the company's first complete capture-and-recreate of a real-world smell.
That demo is the visible tip of a stranger idea. Wiltschko's company is building Olfactory Intelligence - his term for the moment machines learn to smell, the way earlier AI learned to read text and recognize images. Computers got language. They got vision. The nose was the holdout sense, the one that resisted being turned into data.
Osmo's near-term business is fragrance: feeding a creative brief into a model and getting back the first sketch of a perfume formula in seconds, work that used to take a trained perfumer weeks. The longer reach is everything else a nose can do - sniffing out counterfeits, explosives, spoiled food, and the molecular signatures that mosquitoes use to find people.
Computers learned to read and write with NLP, to see with computer vision. We finally taught them to smell with Olfactory Intelligence.- Alex Wiltschko
A sensor reads the volatile molecules coming off a real object - a plum, a flower, a finished perfume.
The Principal Odor Map places those molecules in a coordinate space where nearby points smell alike.
A graph neural network predicts the perceived smell from chemical structure, as reliably as a trained human.
A formula of ingredients is mixed and released - recreating the smell somewhere else entirely.
College Station, Texas. One of the few Jewish families in town, so every summer his parents shipped him off to Jewish summer camp, where he met the fancy kids from Houston and Austin - more confident, more worldly, and, he noticed, better smelling. The cool ones wore Polo Blue, Fierce, Cool Water. Scent was an invisible status symbol, and an eleven-year-old took notes.
By twelve he had saved enough money - earned designing websites and logos - to buy a bottle of Bulgari Black. He describes a single spritz as a drama in three acts: rubber tire, then leather and vanilla, then tobacco. A teenager fell down the rabbit hole of Perfumes: The Guide by Luca Turin and Tania Sanchez, and never climbed back out.
The other obsession arrived in parallel: code. Two hobbies that, in a small Texas town, were each their own kind of weird. He would spend the next two decades pulling them into one career - a neuroscience degree from the University of Michigan, a Harvard PhD in olfactory neuroscience earned in 2016, and a stubborn refusal to let the question of smell go unanswered.
The detour is the interesting part. He did not march straight from the lab into a scent startup. He built machine-learning companies first, learned to ship and sell, then carried those instincts back to the one subject that had nagged at him since middle school. By the time Osmo spun out of Google in 2022, he had roughly eighteen years of scent behind him and a working method for turning a fixation into a product. The perfume nerd and the engineer were, finally, the same résumé.
One summer in middle school, in a moment of insecurity, I became a perfume nerd.- Confessions of a Perfume Nerd
For most of chemistry, you could not look at a molecule's structure and say what it would smell like. Two nearly identical molecules can smell wildly different; two very different ones can smell the same. Smell refused to behave.
At Google, Wiltschko's team trained a graph neural network on thousands of molecules labeled by their odor, and produced what they called the Principal Odor Map - a coordinate space where distance corresponds to perceived similarity. Feed it a molecule no one has characterized, and it predicts the smell about as reliably as a panel of human sniffers. The paper landed in Science in 2023, the foundation the whole company stands on.
Predicts odor quality from molecular structure as well as a trained human.
Using scent prediction to design better repellents against insect-borne disease.
Fragrance is a closed world. A handful of houses control the molecules, the formulas are trade secrets, and becoming a perfumer can take a decade of apprenticeship. Osmo's pitch is to crack that open: hand the model a creative brief and get back the first draft of a formula in seconds, then let a human refine it. Wiltschko frames it as democratization - giving more people the tools to compose with scent, the way software once put publishing and music production on a laptop.
The work blends three ingredients he is careful to name in the same breath: proprietary molecular data, machine learning, and human artistry. The AI does not replace the perfumer. It collapses the slow part - the blind search across a near-infinite space of molecules - so the creative part can happen faster and more often.
When the $70 million Series B closed in February 2026, Two Sigma Ventures led, with Stripe co-founder Patrick Collison among the new names on the cap table. The money came with a heavier org chart: a chief commercial officer out of Amyris and Unilever, a chief operating officer with Givaudan and IFF in his past, and a chief financial officer who had done time at Mercedes-Benz and Tesla. The science had a company around it; now the company was getting an industry's worth of operators. Total raised crossed roughly $130 million.
With AI, we can go from a brief to the first sketch of a fragrance formula in an instant.- Alex Wiltschko
Perfume is the wedge, not the point. Wiltschko's argument is that smell is the oldest sense - the one single-celled organisms used to navigate the world long before eyes or ears existed - and the last one computers cannot do. Teach a machine to read a smell and you get a tool that works wherever molecules drift: a sensor that flags an explosive at a checkpoint, a counterfeit handbag by its glue, food going bad on a shelf, or a disease leaving its trace on a person's breath. With the Gates Foundation, the same molecule-prediction engine has been pointed at insect repellents, looking for compounds that keep mosquitoes - and the diseases they carry - away from people.
It is a long-horizon bet from someone temperamentally built for long horizons. He describes himself as not interested in many things but obsessed with a few, and the through-line of his life is exactly one of those obsessions, picked up at eleven and never put down. Two prior AI companies - Whetlab, sold to Twitter, and Syllable, sold to the biotech Neumora - were proof he could build and exit. Osmo is the one he says he was always going to build, the place where the perfume nerd and the machine-learning researcher finally stopped being two different people. Business Insider put him on its Top 100 in AI; the World Economic Forum named Osmo a Technology Pioneer. The recognition is nice. The plum, reprinted in another room, is the thing that proves the point.
"Scent is nature's original language - one that everyone deserves to speak and understand."
"If I can't make sense of something, it gnaws at me until I do."
"Not interested in a lot of things, but obsessed with some things."
"It will be impossible to catalog all the ways digitizing our sense of smell will positively impact our lives."
"We're digitizing the sense of smell, for human health and happiness."
"With AI, we can go from a brief to the first sketch of a fragrance formula in an instant."