Artificial intelligence (AI) is predicted to reach $ 126 billion by 2025. It appears in all sectors, from healthcare to agriculture, education, finance and shipping. And now, AI has entered the food industry to discover and develop new flavors in food and drink.
In 2018, Danish brewer, Carlsbourg used AI to map and predict the flavors of yeast and other beer ingredients. IBM developed an AI for McCormick to create better spices. And Not co, which produces vegan NotMilks, uses AI to analyze molecular structures and find new combinations of herbal ingredients.
The new company says this new sensory digital footprint for coffee beans will allow roasters and producers to assess quality and taste at any stage of the coffee production process.
The wine-a-fication of the coffee
Felipe Ayerbe, CEO of Demetria, says coffee winemaking is here to stay.
âToday’s coffee drinkers have been exposed and are more aware of the taste and the overall experience they are looking for and are willing to pay more for the experience,â Ayerbe said. “That’s why you see an ever-growing range of possibilities and choices for something that is meant to be a commodity – different prices, origins, roasts, blends, flavor characteristics, preparations – just like wine.”
Ayerbe notes that coffee is still considered a tradable commodity, but the experience consumers get is anything but a commodity. âOver the past 20 years [..], coffee has undergone a premiumisation journey where the most important variable is sensory quality – taste, âadds Ayerbe.
Ayerbe says this revolution in specialty coffees was spurred in part by industry pioneers like Starbuck and Nespresso who improved the taste of coffee around the world.
âWith this sensory digital footprint, we are moving the industry from analog to digital by enabling the entire value chain to [..] measure and manage the industry’s most important variable – taste, “Ayerbe said.” We anticipate that for the first time farmers will not only be able to understand the quality of what they are selling, but also manage their farming practices to optimize their quality, creating an unprecedented level of empowerment for them. “
Reuse sensor technology
The sensor technology used by Demetria has been around for 40 years.
âIn recent years, sensors have become miniaturized, more affordable and can connect to the cloud,â Ayerbe said. “This enables the collection, storage and analysis of huge amounts of data.”
Ayerbe says the company uses wearable near infrared (NIR) sensors to read the spectral fingerprint of green coffee beans. This is because the different colors and wavelengths of the light spectrum react differently to each organic compound present in coffee, representing the entire chemical composition of the beans.
âThen we needed AI to translate the NIR data into sensory language understood by industry,â Ayerbe said. “And, until now, the taste or sensory quality of coffee beans has been determined by suction cups, a manual and tedious process performed by certified industry tasting experts, measured against the standard coffee tasting wheel of industry, âAyerbe said.
With all the data collected from the NIR readings and cupping data, Demetria calibrated the AI ââto match a specific spectral fingerprint to a unique taste profile.
Ayerbe says the biggest hurdle the team had to overcome was figuring out subjective taste identifiers such as body, balance, and aftertaste that aligned with the standardized coffee taster’s flavor wheel.
âThese taste identifiers had to be determined holistically, rather than individually, to establish the true overall flavor profile of the coffee bean,â Ayerbe said. “Additionally, coffee beans from the same sample are not homogeneous, so for a single sample of 300-500 grams of beans, multiple analyzes were required to collect enough data to represent an overall and unanimous flavor profile.”
Demetria performed thousands of scans taking into account the slightest difference between each reading of a wide range of different coffee beans, and the result was a model with a sensory fingerprint unique to Demetria.
Application of sensory fingerprint
Armed with the unique sensory fingerprint, Demetria built a pairing profile for a distinct flavor profile for Carcafe by training their AI and models to use spectral readings and correlating them with tasting analyzes of hundreds of coffee samples.
“The AI ââfor the high value profile required by Carcafe was collected from several hundred green coffee bean samples and measured against quality assessors (cuppers) who submitted data on their scores of tasting, âsaid Ayerbe. “By compiling the analysis of the voluminous suction cups, we were able to match the sensory imprint to this unique Colombian coffee.”
Ayerbe said that after four months of processing the coffee samples, the company created a viable product for Carcafe where the AI ââwas continually recycled with new samples.
âThe biggest technical challenge has been training the AI ââto detect flavor nuances like a cupper can detect, rather than just clear patterns,â Ayerbe said. “Clear profiles must exist in the database, but the full range of shades must also be programmed.”
For the CarcafÃ© profile to which Demetria corresponded, it was necessary to determine a particular sweetness; for example, chocolate is different from caramel which is different from the sweetness of brown sugar.
âSo in the second iteration, we had to define the true taste and make sure that the AI ââcould be specific enough to determine between these very similar but different types of sweetness,â Ayerbe said. âBeing able to identify producers who can exactly match this profile and give farmers the tools they need to be able to cultivate this crop again brings a new level of efficiency and transparency to Carcafe and their customers. “
Ayerbe believes this process removes many variables and unknowns that currently exist in the coffee supply chain.
âIf you can control the process more, you’ll end up with less faulty coffee, which will increase overall uptime,â Ayerbe adds. “It is also important to note that suction cups play a vital role in pattern formation, and this technology is by no means intended to replace their position in the industry.”
Ayerbe says the problem with cupping is that it’s a [..] scarce resource. “We are expanding the ability to assess sensory quality ubiquitously, throughout the value chain, and this is particularly applicable at the producer level where the sucker does not currently exist.”
Ayerbe adds that their technology makes better use of cupping time and allows traders and roasters to be more efficient at understanding who produces the taste, type and quality of coffee they are looking for.