MILAN – Artificial intelligence: a technical tool that has the potential to revolutionize the coffee sector as well, becoming an important aid for every actor along the supply chain. A concrete example of its application, comes with the research of Bruno Pereira de Oliveira, Ph.D. in physics at USP São Carlos and a specialist in coffee tasting, is currently working on the development of a machine dedicated to the goal of detecting quality when it comes to green coffee.
The consequences that may result from this new methodology, will fall on timing, how the raw material is evaluated, as well as its prices.
Being able to conduct an analysis uninfluenced by the human factor, which is inevitably tied to subjective elements, is the priority that guided the researcher to develop a solution that is as objective as possible.
Pereira de Oliveira, how did your research project come about, how long have you been working on it, and what stage is it at?
“The idea was conceived while attempting to adress the issue of quality coffee in the industry, for better elucidation. I aimed to initate my initiate my involvement with coffee during my master’s degree program. The group comprised many individuals from various countries, and upon their arrival in Brazil, our initial iteraction is drink coffee of the origin countries.
Although I cannot pipoint the exact moment starting to the interaction with world coffee, owing to my undergraduate studies in Food Engineering, people frequently engaged me in conversations about the coffee products and the rosting process. During this time, I consumed copious amounts of coffee, as I had not been sleeping much, and coffee became my companion during late-night studies and laboratory work, the interaction with people the other countries, my facination with it grew and search more information about the coffee production.
I began comprehend the sensorial profiles and quality aspects, which proved to be an important turning point in my professional life. This realization occurred during a conference in India in 2018, where I attended to present the results of my Ph.D. stuidies on decontamination processes applied in food industry. However, always the conversation steered towards Brazilian coffee, and during the same conference my mind change. Because I attended a talk by a food scientist discussing coffee blends from around the world.
This inspired me; upon returning to Brazil, I envisioned creating my own coffee brand (Industrial fabric too). Thus, the journey of Espressiata began. My focus was on crafting products that offered unique experiences to Brazilian consumers, moving beyond the ordinary or exclusively specialty coffee (in my mind all the people to consumers a good coffee). To goal it is required a strongly comprehensive understanding of sensory mehrods, markets, supply chains, costs, and the entire coffee production process.
The brand found success in the local market, and after several years, it was acquired by the fourth biggest Brazilian beer company, Império. The challenge then became scalling-up porduction to distribute the same products across all Brazilian states (remenber Brazil have more 35 production differents zone).
At this point, my focus on starting up the line of products and overseeing all aspects of the product development and creating the other news blends products. The blend and the type of green coffee beans to use were clearly defined in my mind. However, when collaborating with others, I encountered varying opinions on the same product. Over the course of one year (2021- 2022), in paralel with my direct research position in Espressiata.
I dedicated myself to researching and developing data science methologies related to coffee quality, processes, and new methods. Despite this effort, I did not see a final product with easy implementation and manipulation. Given my background in Optical Physics from my Ph.D., I contemplated the possibility of developing an optical method to quantify the type of the coffee.
After this year of deliberation and discussion on vairous possbilities (In my mind), I made the decision to drop out my professional position in Espressiata during year 2022 and embarked on a new challeng in 2023: technological development and creation a new brand for it (OptikAI).
In january 2023, our tehcnological readiness level (TRL) was at 1. Currently, we are in the process of migrating to TRL-5, and by the end of this year, we aim to reach TRL-6. I anticipate that in 2024, we will initiate TRL-7 and conduct testing in the field. This transition poses a challenge for us as it requires training people and developing specific skills, because, as the path lies at the intersection of optical and food sciences. Beyond, the necessary the high money to execute all this. ”
How is it possible through light (what light exactly) to distinguish specialty coffee from more commercial coffee, starting from raw?
“To comprehend the basic concepts and phenomena is necessary to move beyond the realm of sensory food science. In the realm of science, there exists a field known as optical physics, which seeks to understand the interaction of light with matter. This field encompasses a wide range of phenomena and technological advancements, resulting in numerous Nobel Prize winners.
Therefore, to evaluate or distinguish the physical-chemical profile fo biological matter, such as coffee beans in this case, is necessary to apply the basics concepts about luminescence, basically, existing fluorescence and phosforescence phenomena that creating especific spectra.
For me, I believie in each this phenomena, will to be divided in more two, inside fluorescence and phosforescence. It is elastic and non-elastic emission spectra (what are these?).
Essentially, this refers to the behavior of the sample in response to interaction with specific wavelenghts of light. Different wavelengths of irradiation yield varying responses in the same sample. However, it is evident to us as scientists that specific samples possess unique characteristic profiles.
I applied these fundamental concepts to assess each type of the coffee (Specialty, Gourmet, riado, rio). Consequently, it became deasible to distinguish betwen samples of green coffee beans without inducing any physical changes and explore new possibilties as well. ”
What chemical characteristics of the bean are examined by this process?
“In this moment not the deconvolution spectrum, however we know that there is a basic structure for each type of coffee and the chemical compounds present in the cellulose structure of the beans represent one specific organized bond between for example sugar, fats, acids and proteins. This conjecture directly influences the quality of the beans, obviously that type of the roasting impacts to emphasize the sensorial tastes and flavors and the other attributes.”
For your work you will have based everything on the elements that make a specialty recognizable: which ones do you think and on which ones did you focus your research?
“I mean, for me is necessary to have the same variable with same equation. The goal in this research is that we development some process, protocol and system to evaluate the aquisition spectrum and the capability of each (basically, determinated the limits up and down). ”
How does a specialty appear different from a nonspecialty?
” I believe that the most significant difference we have observed thus far lies in the concentration of certain specific acids and chemical compounds that possess excited states. These compounds directly influence the quality of the coffee.
I think that the farms have the capability to regulate the generation of these compounds during the operational production from fruit to beans. However, it not yet clear how to market this knowleged these aspects effectively at this moment to effectively contribution.”
What and how much information was used to calibrate the algorithm behind this tool? At what rate of reliability has it arrived at?
“To calibrate the method, we applied the process step-by-step, to control the variable with examples, eletronics, optics, thermals and mechanical with focusing on scalling-up and implementing a rigorous control using the scientific method. It is crucial to thoroughly understand various possibilities before proceeding to the next step.
However, integrating artificial intelligence into the opto-eletronics system poses challenges in ensuring reliability across different methods. While we have achieved excellent results in laboratory-scale experiments, my current focus is on replication these results in the operational prototype. This requires defining the path and expected outcomes every week, as well as determining the subsequent steps.”
Given that with the Sca Speciality coffee association’s new rules, the definition of specialty is increasingly subjective. The new tasting protocol established by the association gives a lot of space and value to elements extrinsic to the raw material such as traceability and social-environmental sustainability, are the parameters you used for your algorithm still compatible with this new paradigm?
“Yeah! I think so, that is possible to contribute with SCA speciality coffee association ( I don’t have contact until this moment), the coffee industry have strongly challenge to solve the method to evaluating the quality without person influence and translate of the qualitative to quantitative results. For me, it is possible to work in these aspects too”
What is the minimum and maximum sample quantity that can be analyzed with your algorithm?
“The idea is to analyze samples ranging from 10 to 30 grams, allowing for evaluation of one sample at a time or continuous operation 24 hours a day, 7 days a week. This approach enables scalability.”
What are the potential benefits of its application for farmers, traders, roasters, and tasters?
” I believe it is possible to contributes to facilities, reproducibility, and, for me, most importantly, impartiality.”
What would be the cost of this tool?
“At this moment, the price has not been precisely defined because it depends on the applications and the necessary automation components that need to be integrated into the device.”
What is the status of patent validation?
“Currently, we have 1 patent deposited in process, 3 awaiting deposition, and several others in the validation stage of results and text construction.”
Are you in contact with the other three groups of scholars around the world who are conducting research similar to yours?
“I don’t have contact with other groups. My primary contacts are in the atomic, biophotonics, and optical groups in Brazil and the USA. However, I am open to interatcting with these groups, as well as with companies.”
