How Tastry "Taught a Computer How To Taste" On AI at Work

Posted by Alyssa Verzino on Mar 1, 2019 5:01:06 PM

Screen Shot 2019-03-01 at 4.37.57 PM

Hearing the classic startup story arc is immensely satisfying. From the discovery of a niche pain point, to constraint-born innovation, to the realization that the solution is scalable to a range of other use cases. Katerina Axelsson, founder and CEO of Tastry, joins us on AI at Work to share exactly that type of story with us. Tastry is a SaaS and insights company whose innovation spans machine learning, analytical chemistry, and flavor chemistry, delivering highly accurate predictions on how sensory-based products - wine, beverages, food, and fine fragrances - will be received by consumers.

So, how did Tastry teach a computer how to taste?

It began with Katerina’s work at a winery’s custom crush facility. She describes that “I would notice that it's very common to have a 10,000 gallon tank of wine, sell half of it to one customer, and half of it to another customer. The same product would go under two different bottles, two different labels, and have different prices. Then, it would receive a different score from the same critics. I thought that there might be an opportunity to create a more objective, insightful system to connect consumers to products.”

To develop this system, Katerina would need to think creatively about how to overcome the challenge of limited data by using alternative methods to collaborative filtering algorithms, which are the most common type of recommendation algorithms, but also require ample data to function.

She points out that sensory-based products require an added layer of complexity in terms of recommendation algorithms. While purchasing behavior patterns on Amazon might be similar for two different people, one might love chardonnay while the other could hate it.

Teaming up with a computer science PhD at Cal Poly, Katerina laid out Tastry’s foundation. They have since patented and validated the ability to track and predict consumer preferences for sensory-based products.

“We delivered a lot of value to a lot of different customers along the supply chain. That's because we broke down the flavor matrix of sensory based products in general. We didn't just train off a dataset of wine. It was about the chemistry that matters to humans,” Katerina describes. Machine learning is instrumental for this type of application, as there are countless inter-dependencies among the thousands of compounds in a single product. In this way, Tastry’s work isn’t replacing a sommelier. “We’re doing something humans can never do, which is identifying these inter-dependencies and complexities in a product,” says Katerina.

Large flavor and fragrance companies are keen to work with Tastry, because of how this startup is fundamentally changing the product development cycle in a way that significantly mitigates the risk of failure. Typically, when a new product debuts, companies have to wait to see how it performs in the market. On average, there’s an 80% failure rate for a new product that’s launched into the market. With Tastry’s technology, Katerina explains, “we can simulate how it would perform and make modifications before it hits the market as well. No one was doing that before.”

As she contemplates what the rest of 2019 holds in store for Tastry, Katerina references IBM beating Kasparov at chess, saying “I would like for Tastry to identify various wines alongside a world class sommelier.” It’s a competition we can look forward to watching in the near future.

Tune in to this episode of AI at Work to learn more. And, be sure to subscribe on iTunes, Spotify, or Google Play and share with your friends.