UNIVERSITY PARK, Pa. — When Jason Cohen, CEO of Analytical Flavor Systems, talks about the work his young company is doing, you can hear the passion in his voice.
The start-up, launched in 2012 while Cohen (B.A., Political Science) was an undergraduate at Penn State, stemmed out of his research on the flavor profiles of tea, conducted under professors Ryan Elias and Joshua Lambert in the Department of Food Science.
But Cohen had a data problem. He just couldn’t recruit enough tea tasters. “So I started gathering data on coffee tastings,” Cohen recalls, “but I still couldn’t get enough tasters. Then I started holding beer tastings — and I no longer had a data problem.”
Around this time, Cohen and co-founder John Dori moved their research to Penn State’s College of Information Sciences and Technology (IST). Working alongside John Yen, they developed a software application that streamlines data gathering and analysis.
As Cohen talked to coffee roasters and beer brewers about their research, he realized he had a service that he could sell. Analytical Flavor Systems was born. He made the strategic decision to postpone his integrated master's research in food science and machine learning in IST in order to develop his business.
The concept of Analytical Flavor Systems is simple: Trained tasters use a smartphone to record reactions to 25 different factors as they sip from a new product batch. The software, dubbed Gastrograph, collects an additional 600-1,000 pieces of data, including date, time, altitude and temperature. Everything is added to a database of tens of thousands of tastings, and then analyzed.
Over time, the researchers developed strong predictive models that help drink makers identify strengths and weaknesses in a new brew. More importantly, Gastrograph recognizes when the production process missed the mark.
“Because the ingredients in food and beverage production are agricultural,” Cohen explains, “they vary in flavors, and can lead to inconsistencies.” That’s where Gastrograph comes in, providing early detection and correction during the production process.