On a recent Tuesday, Alex Wiltschko invited me to his company’s lab in Kendall Square to sniff a few synthetic fragrances. Wiltschko owns a black plastic case full of vials, like a traveling scent salesman. He dipped a thin strip of paper into each vial, and handed it to me. One resembled a delicate lily of the valley, the next a powerful jasmine, and following that a bright red berry.
Wiltschko describes himself as “a scent-obsessed individual” — one who collected perfumes as a kid, and later earned a doctorate in olfactory neuroscience at Harvard — and now he is building a startup company that is equally obsessed. Osmo has 17 employees, $60 million in funding, and a drive to get really good at using machine-learning algorithms to understand why natural things like carnations or coconuts smell the way they do, and produce a realistic chemical clone.
Why would a scent-obsessed founder and his company attract that kind of funding? One reason is that naturally-occurring fragrances are under pressure for safety and environmental reasons. “They may not be biodegradable when you wash it down the drain,” Wilstchko said. “Skin sensitivity is another issue. Some may be carcinogenic.”
One example: last year, the European Union banned lilial, a floral fragrance, after research raised concerns about its impact on fertility. Kelly Kovack, CEO of BeautyMatter, a beauty industry trade publication, notes that the production of ingredients like sandalwood oil or vanilla, can sometimes lead to deforestation in developing countries — and that many consumers prefer to purchase products that carry a “sustainably sourced” label.
Wiltschko said that the people who design fragrances for a new shampoo, lotion, or perfume have a palette of perhaps 2,000 fragrance-related ingredients, and as many as half of them “might not be around in 10 years, because the standards are rising.” Osmo wants to supply synthetic replacements.
The company is also working to develop what you might call “functional aromas,” such as substances that repel mosquitos. “We’ve already found eight or so ingredients that are more potent than DEET,” said Wiltschko. That’s the chemical used in many of today’s insect repellents. Wilstchko said the company is conducting tests with a partner in the Netherlands. And he hinted that there are other benefits to endowing computers with the ability to understand smells as animals and humans do: rats can identify tuberculosis by smelling a sample of phlegm from an infected person’s lungs, he pointed out, and dogs can detect various types of cancer.
Osmo’s lab is a fragrant place. When I arrived, it smelled to me like cotton candy was being spun on a carnival midway. Wiltschko described it as burnt sugar, but said this particular synthetic smell is a bit off target: they were aiming for butter.
Osmo doesn’t plan to build hardware of its own to synthesize or analyze smells. Rather, it has created a neural network — software that can mimic the way the brain finds connections and patterns — that has been trained on a vast set of information about the odor that different molecules emit. That allows Osmo to ask the software to design entirely new molecules, or combinations of molecules, to produce a particular smell. But the synthetic doesn’t always smell exactly as it should — like burnt sugar rather than butter — so human noses are needed for quality control.
Wiltschko uses a typewriter keyboard as a metaphor for smells. Today, he said, “we don’t have enough keys on the keyboard to emulate the smells found in nature.”
David Edwards is an entrepreneur and researcher who has worked on scent-related projects for most of the last decade. He sees value in Osmo’s approach for creating “improved versions of natural fragrances,” but Edwards also says that advances in the understanding of olfaction have “made it possible to conceive of a new field of therapeutics,” where specific scents, perhaps combined with pharmaceuticals and then inhaled, might treat conditions like anxiety, sleep disorders, or addiction.
Other companies have been sniffing around for opportunities in this arena as well. Boston-based Ginkgo Bioworks has worked to develop synthetic fragrances for customers such as Givaudan, a flavor and fragrance manufacturer based in Switzerland. The company has worked on three customer projects in that industry, said Christina Agapakis, Ginkgo’s head of creative and marketing. And an art project that Ginkgo was involved with, “Resurrecting the Sublime,” sought to synthesize the fragrance of extinct flowers. (It is currently on display at the MIT Museum in Cambridge.) But Agapakis said that working on fragrance development, while “still something that is important to us,” does not represent a significant chunk of Ginkgo’s $478 million in annual revenue.
Osmo’s $60 million comes from GV, a venture capital firm affiliated with Google, and the Bill and Melinda Gates Foundation, among other investors. (Some of Osmo’s foundational technology was created when Wiltschko worked for Google, as part of Google Brain, an AI-focused research group.) The company rents lab and office space in an MIT-owned incubator called The Engine, which opened last year in Cambridge. In an earlier era, the building was home to the corporate offices and chemical labs of Polaroid, the company that pioneered instant photography.
Wiltschko collects cameras, and one of them is a 1960s-era Polaroid Land camera given to him by an aunt. When he first took possession of it, he noticed that there was an undeveloped piece of film trapped inside. He pulled it out, and as it developed, he saw a picture of one of his cousins, taken four decades prior.
In the same way that our ability to see things and remember them was transformed by photography, first with chemicals on paper and later with pixels, Wiltschko predicts that the same thing will play out with our sense of smell. The digitization of our sense of vision took more than a century, he pointed out. “I think we can condense that into 10 years,” he said.
Scott Kirsner can be reached at firstname.lastname@example.org. Follow him on Twitter @ScottKirsner.