We are building the Large Taste Model (LTM): an AI recommendation engine built on sensory science, designed to understand, predict, and personalise your next favourite drink.
Speak to us at hello@largetastemodel.comArtificial intelligence has successfully mapped language, audio, and vision, creating rich, high-dimensional representations of human knowledge. Yet, the physical world of chemical perception remains largely untouched.
A lab test can tell you the exact molecular makeup of a drink, but a machine can't actually taste. It can't feel the weight of a full-bodied wine on the palate or the refreshing crispness of carbonation.
Flavour only comes to life when someone takes a sip. That's why we aren't just cataloguing chemical data. To build a recommendation engine that actually works, we are focusing on how humans translate those ingredients into real enjoyment.
We design our system to mirror the human palate. By mapping how our senses translate chemical makeup, texture, and temperature into a single, cohesive taste, we help technology learn how we actually experience flavor.
Our model matches molecular profiles directly to real human vocabulary—learning the exact connection between chemical compounds and the words we use to describe them.
It’s all about capturing the personal, rich, and real-world experience of taste.
We're proving the model first in wine; a domain with an unusually rich corpus of human tasting language and enough structure to test whether predictions match real palates. The same representation generalises: spirits, coffee, food pairing, and beyond.
Use AI to find empty spaces on the flavour map, uncovering entirely new combinations of compounds that are mathematically balanced.
Accurately forecast how individuals will perceive novel blends, ingredients, and chemical combinations before they are physical.
Create a universal language for taste, allowing precise translation of complex sensory profiles across different physical mediums.