Deep Learning for Coffee Analysis
Can a machine judge coffee objectively? Ryan Koes ’26 explored that question with sensors and machine learning, co-advised by Prof. K. Wakabayashi (Chemical Engineering), Prof. Alan Marchiori (Computer Science) and me.
He started with classical models (support vector machines, decision trees, logistic regression) that use measurable properties such as pH, turbidity, dissolved solids and color to tell coffees apart. He then moved to deep learning on cyclic voltammetry: electrochemical signals from screen-printed electrodes that respond to compounds like caffeine and chlorogenic acids.
The work became Ryan’s honors thesis, Deep Learning Approaches for Voltammetric Analysis of Coffee, which he successfully defended at the end of spring 2026.
Ryan’s love of coffee went beyond the lab. In my XR course he built an XR barista simulator that teaches you to pull an espresso shot.
Back to top