Vectorized vineyards: Big data in wine making

Suggestions from a silicon sommelier If you’ve ever walked into a wine store seeking an offering for a dinner party, you know how crowded the market can be. Last year, wineries applied to sell about 126,000 unique labels. “You don’t see that in ketchup,” says industry expert Katerina Axelsson. Axelsson hopes to help prospective purchasers choose amid this dizzying array of options by using wine chemistry, consumer preferences, and a hefty dose of artificial intelligence. Before founding her company Tastry , Axelsson worked in the analytical laboratory of a winery. “The problem is, machines don’t look at chemistry the way humans do,” Axelsson tells Newscripts. For a spectrometer, ten micrograms per liter of benzaldehyde is ten micrograms per liter of benzaldehyde . But for a human, whether that compound is noticeable depends on its context. Another compound might mask benzaldehyde or might increase its potency—or might make it register as