How Osmo is digitizing smell with Google Cloud AI technology

Smell is a complex and profound sense. Our noses are powerful enough to detect Parkinson’s disease earlier than any diagnostic, and dogs, bees, and other animals can sniff Alzheimer’s, COVID-19, and even cancer. And yet, the olfactory system is still poorly understood, because we have been missing something the other senses have — a map. Color has a map, represented by the axes “red,” “green,” and “blue.” Sounds are mapped by frequency, from low- to high-frequency. Nothing similar has existed for odors.

Osmo is working tirelessly to change that, with the help of Google Cloud.

At Osmo, we believe digitizing smell can unlock countless new possibilities - from creating the next generation of more sustainable aroma molecules for perfumes, shampoos, and other everyday products to preventing mosquito-borne diseases like malaria and dengue fever by predicting molecules that could repel insects. Before Osmo was its own company, I spent years working with fantastic colleagues to validate its digital olfaction research at Google Research on the Brain team. And now, with Google Cloud’s AI technology, we’re tackling the challenge of mapping our sense of smell in order to improve the health and well-being of human life. 

AI is making it possible for us to undertake cataloging the incredibly vast world of scents, made up of an estimated 40 billion molecules that are likely to exist and have an odor, of which only 100 million are known. To comb through this immense amount of data, we are utilizing a variety of Google Cloud AI solutions, including Vertex AI for training. Vertex AI makes it easier for our developers and engineers to quickly and easily train AI and machine learning models and simplifies the process of building, training, and deploying machine learning models at scale in a single environment, helping us more efficiently discover patterns and make better predictions. 

After training our models on Vertex AI, we run the models across a large library of molecules using Dataflow to write the results in BigQuery and Bigtable. We’re using Dataflow to run models on these large datasets in hours instead of weeks, with a single machine. Dataflow enables us to process large amounts of data quickly and at scale, with a serverless approach that reduces operational overhead and enables our team to focus on their important work.

BigQuery collects the data of billions of scent molecules to accelerate our analytical insights, improves Vertex AI’s machine-learning models, and ​​manages massive datasets at scale (a table in BigQuery of 40 billion molecules is 15 TB). BigQuery enables us to ask questions of datasets of billions of rows very quickly; getting the same results from other tools would take hours instead of minutes. Ultimately, the scalability of Google Cloud’s tools has powered our efforts to find new molecules no one has smelled before. 

“Google Cloud is committed to supporting innovative, fast-growing startups such as Osmo and helping them expand within our open data cloud ecosystem. We are honored to be helping Osmo’s journey toward pioneering digital olfaction.” – Gerrit Kazmaier, VP and GM of Data Analytics and Looker.

As a start-up, it was important to us to choose a data cloud that would enable us to scale in any direction we need to, without concerns about lock-in down the line. One of the main advantages of utilizing Google’s data cloud is its open data ecosystem. For example, BigQuery provides built-in machine learning that works across clouds and scales to support running analytics on our various data types, so we have the flexibility to integrate with tools outside of Google if we ever need to. 

With the help of Google Cloud, we are building the foundational capabilities that will ultimately enable  computers to do everything our noses can do. With the power of world-class technology and our world-class team, we are creating a world where digitized smell will be a critical tool to help us build more beautiful and ecologically-friendly fragrance experiences, detect diseases earlier, track pandemics faster, grow more food, ward off insects, and much more that will transform human health and build a better future.