Research
Publications
Aug 22, 2026 · Current Research in Food Science
Predicting food taste with bound-driven optimization
Pagkratios Tagkopoulos, Dimitris Sfondilis, Ilias Tagkopoulos, Tarek Zohdi
This paper treats a recipe the way a materials scientist treats a composite: a mixture whose overall properties are bounded by the properties of its parts. The authors applied classical mixing bounds from material science to 70 recipes and compared the predicted sweet, salty, sour, bitter and umami intensity against how people actually rated them.
Apr 8, 2025 · Presented at COLM 2026
SkillFlow: Scalable and Efficient Agent Skill Retrieval System
Fangzhou Li*, Pagkratios Tagkopoulos*, Ilias Tagkopoulos(* equal contribution)
SkillFlow is a retrieval pipeline that lets an agent find the right skill in a very large library, about 36,000 community-contributed skill definitions indexed from GitHub, by narrowing in four stages: dense retrieval, two rounds of cross-encoder reranking, and a final LLM selection.
Oct 4, 2026 · Under review at npj Science of Food
Integrated omics reveals actionable drivers of bioactive variation in US milk
Cheng-En Tan, Mariana Barboza, Pagkratios Tagkopoulos, Shanghyeon Kim, Lukas Maximilian Masopust, Ivor Prado, Muhammad Adil Salim, Fangzhou Li, George Berdovskiy, Ilias Apostolakos, Brandon Invergo, Karen M. Kalanetra, Danielle G. Lemay, David A. Mills, Armin Oloumi, Cheng-Yu Weng, Carlito B. Lebrilla, Xuan He, Carolyn Slupsky, Oliver Fiehn, Ibuki Kusumoto, Ameer Y. Taha, Yu Wang, Daniela Barile, Alexis Davis, David E. Olson, Justin B. Siegel, Ilias Tagkopoulos
The study measured 60 retail milk samples from ten US regions in seven ways at once, from genes and RNA to proteins, peptides, fats, sugars and small molecules. It identified 6,714 compounds and gave absolute concentrations for 5,288 of them, most of which had no such estimate in existing milk databases.
2026 · Under review at Briefings in Bioinformatics
Disease-level drug approval forecasting from trial-agnostic public data
Samir Townsley, Pagkratios Tagkopoulos, Ilias Tagkopoulos
The paper forecasts drug approvals at the level of a disease rather than a single clinical trial, using only public data that does not depend on the details of any one trial.
Feb 5, 2026 · bioRxiv
Design to Data for mutants of β-glucosidase B from Paenibacillus polymyxa: Y333F, A88E, L219Q, A408H, Y173L, E340S, and Y422F
Alexandra Maduros, Leila Farinsky, Pagkratios Tagkopoulos, Ashley Vater, Justin B. Siegel
Design to Data is a UC Davis program that builds that dataset one mutant at a time. This paper adds seven single-point mutants of β-glucosidase B, each designed in Rosetta, then made, purified and measured at the bench.
Laboratories
Mar 2024 to present
USDA/NSF AI Institute for Next Generation Food Systems (AIFS)
Research Assistant, University of California, Davis
Developed SkillFlow, a framework for AI agent code evolution and skill retrieval. Designed orchestration and state-machine architectures for AI assistants, and assisted in the development of FoodAtlas, an AI-driven knowledge base on food, chemicals and health effects. Also contributed to an integrated omics study of US milk and to a model that forecasts drug approvals by disease.
Mar 2024 to present
Siegel Lab, NSF Design2Data
Research Assistant, University of California, Davis
Protein engineering of β-glucosidase B. Ran the end-to-end pipeline: computational Rosetta design, experimental mutagenesis, selection, purification and activity measurement.
2025 to present
Zohdi Group, Department of Mechanical Engineering
Research Collaborator, University of California, Berkeley
Worked with Professor Tarek Zohdi on whether the taste of a dish can be predicted from its ingredients, using mixing bounds from materials science. The work became the paper "Predicting food taste with bound-driven optimization".