Where biology meets machine learning and yields unexpected insights
It all started as a kid growing up in a small town in Texas. Alex was “obsessed with scents” since he was very small, and devoted himself fully to his hobbies of collecting perfumes and being a computer nerd, which, as he puts it, “was not exactly a recipe for success in the popularity contest.”
Regardless, these obsessions stayed with Alex and guided his early life decisions, leading him to medical school to study olfactory neuroscience at Harvard University. There, he trained under Sandeep Robert Datta, who trained under Richard Axel, the winner of the Nobel Prize for his discovery of the olfactory receptors. Another close mentor was Ryan Adams, whose mentor was Geoff Hinton, who many credit as the inventor of deep learning. His career since his training has been about fusing those two worlds: biology and machine learning–specifically at the intersection that is now known as olfactory neuroscience and machine learning, or “machine olfaction.”
Alex took a detour after graduate school, founding and selling two AI companies–one to Twitter, and the other to a biotech called Neumora–before joining Google Brain about six years ago, where he led a couple teams before founding the digital olfaction group.
The group achieved a breakthrough: “the ability to predict what a molecule smells like more accurately than a person can describe it,” says Alex. They conducted a first-of-its-kind study, training a type of AI known as a graph neural network to predict what a compound will smell like to a person, based on the chemical features of odor molecules, and the computer model assessed new scents as reliably as humans.
Alex and the team at Google Brain shared their study results with Christophe Laudamiel, one of only six-hundred people in the world who can call themselves a perfumer (and now an advisor to Osmo), who helped them to realize that many flavor and fragrance companies–developing anything from shampoos to lotions and candles–were interested in these molecules. They soon determined that the best way to scale this opportunity was as a spinout new venture, founding Osmo in 2022 with co-leads and co-founders Lux Capital and Google Ventures (before later bringing on a list of other supporters, including the Bill & Melinda Gates Foundation, Two Sigma Ventures, and more).
Much of the inspiration for Osmo is born out of revolutions in drug discovery and pharma over the past ten years: using machine learning to predict how molecules will behave or what properties they will have before even creating them. Osmo sets out to extend, specialize, and apply cutting-edge data science-driven drug discovery methods to make sense of the world of scent, training neural networks to predict what a molecule will smell like.
So what does that prediction look like?
It turns out that there is a vocabulary for describing smells, with hundreds of terms that can be organized and mapped using AI, effectively creating a new taxonomy for scent, or what the Osmo team refers to as “a map of odor.” This map is a critical discovery because every other sense has a map–like how color has RGB (three numbers that communicate what any color is and how to mix them), or how audio has low to high frequency.
This map will not only produce new knowledge, but will also bring exciting new scents to people, create more sustainable alternatives to scents that are created from nature today, and even help solve big challenges such as disease detection.
When a very strange bird enters a centuries-old fragrance market
With Osmo tackling the fragrance industry first as their beachhead market, I can’t help but wonder what the response from perfumers would be to this challenger business model, built entirely using AI from the very start. Though Alex believes “the response has been one of high interest and curiosity,” he acknowledges that “overall, we are a very strange bird, a very new bird in this world. And we recognize just how different we are, and the depth of the tradition that we’re stepping into.”
Osmo’s fragrances will be designed in a way that not only optimizes the latest advancements in deep tech, but also in sustainability. “The scale at which you have to farm and harvest living things to only get one drop of fragrance can be astounding,” says Alex. Further, “not all molecules or ingredients are created equal with respect to how they function after you’re done with them.” Osmo is building sustainable synthetic substitutes that have the same odor character to the original molecules, but that can be more advantageous for the planet. Alex believes that Osmo can help us to “enjoy flavors and fragrances that create milestone memories in our lives, with less guilt.”
Colin Beirne, Partner at Two Sigma Ventures who acted as our lead investor for Osmo’s Series A, says, “We were smitten with Alex and this idea from the first minute. It’s a dream combination: a founder with top percentile expertise and obsession about a topic, with near-term markets they can attack, and a long-term vision that could make a big dent in multiple huge societal challenges. It won’t be easy, but it’s tremendously exciting work.”
Why Osmo’s map of odor means big things for the planet
On the broader opportunity for Osmo, Alex shares:
The first step Osmo is focused on right now is designing molecules that smell in a specifically intended way. Early applications beyond flavor and fragrance include the creation of molecules that can help prevent insect-borne diseases. In working towards that, Osmo has already created a new kind of molecule that can propel mosquitoes, conducting this research in partnership with the Bill & Melinda Gates Foundation.
Where is Osmo’s staircase of scientific invention leading to?
“Where we’re going is taking osmographs. In the same way we can take a photograph of a loved one’s face or a beautiful landscape on our vacation, we should be able to take an osmograph and capture our scent memories that really do define touchpoints in our lives,” shares Alex.
And while that is a very difficult technical challenge, most of us probably don’t know that there are a couple thousand people in the world that can already do this through a very laborious, manual process involving a sensor that turns physical atoms into bits; a map that interprets, stores, compresses and transmits scent like a JPEG, and a printer that translates bits back to atoms. Osmo’s vision is not to invent the technology, but instead to integrate what is being pioneered elsewhere with the map of odor that Osmo has built, which is critical to this larger vision.
Even if this vision may seem to some as even more surreal than Nikolai Gogol’s short story, “The Nose,” where the protagonist’s nose leaves the face to take on a life of its own–Alex assures us that its arrival is an imminent future.
“Just like the photograph that went from something that was extremely niche 250 years ago, to something that millions and then billions of people could do effortlessly. This is happening and will happen for osmographs, which will include things like food that you’re instagramming anyway, but also…”
Who is the scent obsessed band of misfits behind this breakthrough?
The eclectic team of experts behind Osmo includes the person who built Spotify’s song recommender you likely use day in, day out; people who have built autonomous driver systems; people who have run product organizations at large agricultural tech companies, with deep experience in hardware prototyping and manufacturing; several neuroscientists specializing in molecular or cellular neuroscience and human psychophysics, and more.
And they are expanding on these capabilities to include new hires that bring expertise in small molecule development for drug discovery, as well as analytical chemists that specialize in operating large chemical sensors. Osmo sees this mix of skills as essential for arriving at the right blend of expertise to enable the first osmographs.
Alex encourages anyone out there who is “secretly scent obsessed, and thought that this vision was possible–giving computers a sense of smell,” to consider joining Osmo.
We can’t wait to continue to support Osmo for the journey.