Hi, my name is Alex Wiltschko. And if you remember only one thing about me, it’s that I’m obsessed with smell.

I grew up in a small town in Texas, where neither computers nor perfume were popular. I inherited both my Dad’s passion for computers and his sensitive nose, and became obsessed with software and scent, becoming both a computer nerd and — oddly enough — a perfume collector. I knew a few kids growing up who loved computers as much as me, but never did find a Texan as scent-obsessed as myself. Given my interests, I wasn’t exactly in the running to be the most popular kid in high school. But I couldn’t help but follow my passions. I ended up studying olfactory neuroscience all the way through graduate school … and learned how little we still know about our sense of smell.

Halfway through graduate school, my Dad got sick. It first manifested in ways we didn’t catch early on. He spoke less and less. He got little tremors in his arm that he couldn’t control. And then he had a seizure while on vacation. After some scans, doctors found a set of tumors in his brain, some the size of golfballs. After more tests, he was diagnosed with glioblastoma multiforme. We’ll never know how long it was growing before we caught it. It could have been weeks, months, or years. I left school to take care of him until he passed away 6 months later. It was the hardest stretch of my life, but I’m grateful to have been there for him all the way until the end. He was calm, gentle, funny, and I still miss him deeply.

It took me a while to process all that happened. Years later, I returned to those memories with a scientist’s mind, and I learned that some cancers have a smell. Diving deeper, I looked for a pioneering research paper. Instead I found a slew of papers, each with its own strengths and weaknesses, that collectively posed the simple question that would come to guide my life’s work: Why can’t computers understand smell? Why can’t they be constantly monitoring the world around us for opportunity, danger, and disease? Why do we lose people we love before their time when we know we can do better? The same sense of smell that brings great joy to our lives can also be used to prevent pain and suffering. Giving machines a sense of smell means giving everyone a chance at a better life, and we have to mount an all-out attack on this challenge.

My obsession took hold, but I was on a meandering path. I co-founded a company called Syllable commercializing my PhD work analyzing animal behavior with computer vision, which was sold to Neumora, a big biotech company. I co-founded a second company called Whetlab that was an early Machine Learning as a Service company, which we sold to Twitter. At Twitter, I learned industrial-scale machine learning, and then I moved to Google Brain, doing more of the same. After a year of pure-computer science projects, I felt it was time to dedicate myself full-on to the historic challenge of digitizing our sense of smell – a journey that would start at Google and continue with the launch of a brand new company, Osmo, dedicated exclusively to this mission.  

I can’t wait to tell you all about it.

Alex

Chapter 1 – Introducing Osmo

It’s time for humans to give computers a sense of smell.

When computers learned to see and to hear, our species changed forever, and for the better. Teaching them to smell can transform human health and well-being.

Smell is our oldest sense, and it’s wired directly into our emotions. Our own sense of smell is phenomenal, and still mysterious. Our noses can detect Parkinson’s disease earlier than any diagnostic, sniff Alzheimer’s, COVID-19 and cancer. Why can’t our computers? Computers haven’t understood smell because we’ve had no map of it (imagine trying to build a camera without knowing about RGB, or a microphone without knowing about frequency).

Until now. For the first time in human history, we have a map of odor.

Today, I’m pleased to announce Osmo –  a company that intends to use this map to digitize smell to improve the health and wellbeing of human life.

It’s a lofty mission – one that will take Osmo’s world-class team, multi-disciplinary approach and unprecedented combination of hardware, software, data and capital ($60M in funding at launch) to achieve.

Where we’re starting: Flavor & Fragrance

In time, digitizing smell will help us detect diseases earlier, track pandemics faster, grow more food, catch food spoilage before it harms, ward off insects … and much much more. As a startup, Osmo will be starting in a space where we can create value today while building a foundation for future diversification: the flavor and fragrance industry.

You might not know about this industry, but there are a small group of companies that make every smell in nearly every product you experience. They’re called Flavor & Fragrance houses, and they’re responsible for making many of our memories, from childhood to adulthood — the smell of fresh laundry, the taste of your favorite snack, the perfume of your high-school sweetheart, the cleaning spray used on the dining room tables at your summer camp. These are all carefully constructed products, made from ingredients designed specifically for their smell.

There’s a problem, however: many of these ingredients need to be better. They need to be more biodegradable and sustainable, so they don’t endanger our environment. They need to be safer on and in our body, so that they don’t make us feel unwell. And still, they need to smell fantastic, work in a variety of products like perfumes, detergents and hand-soaps, and be affordable to produce. This problem is getting harder over time, not easier, and many ingredients that perfumers and flavorists depend upon are being removed from the market because they’re not up to our increasing standards. F&F companies are continually making new ingredients to replace those that are being lost to regulation or disruptions in the supply chain.

The design process for making new molecules, however, has been entirely manual. A fragrance chemist starts with a molecule that they know smells great, modifies it slightly, and tests it. Teams of fragrance chemists might do this thousands of times per year, and find just one or two that satisfy all of the requirements I mentioned above, making it good enough to put on the market. We’ve already built an AI system that can design new molecules to precise specifications, and created molecules that smell great in the hands of master perfumers, industry experts, and double-blind odor panels. Osmo’s first steps to giving computers a sense of smell is to use our map of odor to build better, safer, more environmentally-friendly fragrance ingredients. We have worked for many years on artificial intelligence and machine learning for olfaction, and have developed deep expertise in this space. But, we approach the flavor & fragrance industry with a deep humility; many of the best companies in this space have been working on these problems for over a hundred years, and have adapted to a changing world many times over.

While our entry point is flavor & fragrance, our ultimate goal in giving computers a sense of smell is to improve human health and well being, and it’s going to take a long time to achieve this goal. We will never lose sight of this. Applying our technology and expertise in the flavor & fragrance industry is an exciting and impactful first step in building a business meant to last longer than a lifetime.

How we’re building it

In pursuit of this vision, Osmo is bringing together a founding team of world-class neuroscientists, machine learning experts, psychophysicists, hardware and software engineers, data scientists, analytical chemists and industry experts: Chase Buchholz, Mike DeTienne, Rick Gerkin, Jon Hennek, Brendan Lehnert, Rohinton Mehta, Harry Pellerin, Wesley Qian, Rich Whitcomb, and Jake Yasonik. I’m honored to be on this journey with them.

Together we’re building on work that started at Google Research, where I led a digital olfaction team that used cutting-edge machine learning and careful laboratory experiments to establish the foundations for a map of smell. In building and validating this map at Google, we achieved unexpected scientific milestones — superhuman scent prediction, the development of new insect repellents that could save lives, and a deeper understanding of the biological forces of olfaction itself.

We decided the best path for growth and investment for our mission was through a separate startup, which would give us the speed and flexibility to tackle the unique challenges ahead of us. We co-founded the company with Lux and GV – gaining incredible partners in Josh Wolfe, co-founder of Lux Capital, who had also been on a decades-long pursuit of digital olfaction, and Krishna Yeshwant, who co-leads the life sciences team at GV. We’re excited to build this company with Josh as he joins our board, and with Krishna as a board observer. Serial entrepreneur and investor Andy Palmer joins as our founding independent board member.

With Lux and GV as co-leads, we’re pleased to announce a $60M Series A funding round for Osmo. Other fantastic investors and funding institutions join the round, including Arena Holdings, the Bill & Melinda Gates Foundation, Moore Strategic Ventures, Exor  Ventures, Two Sigma Ventures, and the Amazon Alexa Fund; individual investors include Hugo Barra, Soumith Chintala, Jeff Dean, Henry R. Kravis, Rich Miner, and Thomas Reardon. This is an amazing group of funders and investors, and we couldn’t be more thrilled to be taking this journey with them.

While we’re not looking for more funders at this time, we are looking for “Osmonauts” to power and guide our journey. Our simple mantra is “be kind, get stuff done.”If you’re passionate, have a love of scent, love to learn, and have a deep expertise in an area the company  needs, consider joining us. We’re looking for a head of chemistry, experts in chemoinformatics, data scientists, DevOps engineers, ML engineers, product managers, analytical chemists, and chemical lab managers. Check out our Careers page for positions we’re trying to fill.

Where we’re going

Fulfilling our mission of digitizing smell to improve the health and wellbeing of human life will take a while. At Osmo, we are building the foundational capabilities towards enabling computers to do everything our noses can do. We will be successful when people live longer, happier lives because of digital olfaction. This is a challenge for our society, and for our species as a whole.

It took a hundred years to digitize vision, starting from the advent of popular photography. There was a time before the photograph when the only record of the visual world was in carefully crafted paintings or our own minds. Just like the photograph allowed us to capture light, we imagine that the osmograph will capture the smell and taste of the world. Scents make memories, and memories make up the stories of our lives. Scents tell us if we’re healthy or sick, and knowing that information early can save lives.

We’re on a long road, but we’re taking our first steps today. It took 100 years to digitize our sense of sight, and we will digitize the last human sense, our sense of smell, in a fraction of the time.

Chapter 2 – Building a Map of Odor

What does it mean to digitize a sense?

Look at how it happened with vision. Before everything, we made a map. RGB is a map of color. Three numbers — the amount of red, green, and blue — can specify any color. With this map, we discovered the cell types in our eyes that perceive these colors. When we built silicon sensors that are sensitive to light, we used this same map of color to build the chips, and to interpret the data coming off of the chips.

For sound, we discovered that sound is organized along a single axis, from low- to high-frequency. The deep bass of 20Hz all the way to spoken language at around 2000-10000 Hz, to inaudible at 20,000 Hz. Frequency is a map of sound, and it helped us discover how a structure called the cochlea in our inner ear lets us hear, and it enabled us to build a microphone.

So, to digitize sight and sound, we needed a map of sight and sound. But what exactly is a map of smell?

This question had gone unanswered for 100 years or more. At Google Research, where I founded and led the digital olfaction team, our first job was to define the problem. We went with a straightforward definition: could we predict what a molecule will smell like? This is called the Structure-Odor Relation (SOR) problem.

I had studied this a little while I was in academia, but at Google, massive breakthroughs had been made in the field of machine learning for chemistry. The breakthroughs had been applied to building better drugs, but I thought they might help crack the SOR problem. I was fortunate enough to bring on an amazing group of researchers and engineers to take dead-aim at this problem.

A map of smell

It’s reasonable that science hadn’t yet developed a map of smell. For sight, RGB is a 3D map of color because there are 3 types of receptors that detect color in our eyes, each corresponding roughly to red, green, and blue. Smell on the other hand, uses hundreds of receptor types. Science is good at building maps that are 2D or 3D and can fit on a flat sheet of paper or a globe. We needed to wait for the advent of machine learning to build maps – like those needed for smell – that have dozens or hundreds of dimensions.

High-dimensional maps can be built with neural networks, which can capture complex patterns in datasets. For instance, Google Image search works by building a map of all images, and converting your text search into a coordinate, and finding all images nearby that location on this high-dimensional map. It’s the same for Google Translate, which is based off of a large neural network that has a representation of all human language, and that’s what lets you translate from English to Portuguese, from French to Mandarin, and back. All human languages are represented in a common map. My team at Google Research used cutting-edge machine learning to build a map for smell.

To solve the SOR problem and predict what a molecule smells like from its structure, we made a major breakthrough using a relatively new kind of machine learning called Graph Neural Networks. We spent years validating our work at Google Brain, some of which we just shared with the world in a blog post. We built new molecules no one had ever smelled before and predicted them with superhuman accuracy. We built molecules that smell bad to mosquitoes (e.g. insect repellents) and are more potent than DEET in human trials (this could save lives). And we discovered that species as diverse as humans, mice and insects may share the same odor map, just like our eyes all use RGB. Why is that? The light from the sun has been the same to living things for billions of years, so all of our eyes evolved to take advantage of that. Over evolutionary time, life all breathed the same air, so it’s reasonable to expect that we developed the same map. The scents in air we breathe are made by other living things, so perhaps our sense of smell evolved to smell the building blocks of life.

Not only does our odor map work to build new and valuable molecules, it tells us something fundamental about biology and life. At Osmo, our aim is to build on our learnings from Google Research and use our map to discover ever more valuable ways to save and improve human lives through smell.

‍