Jan Frick

Jan Frick is a PhD candidate at BHT under the supervision of Prof. Dr. Alexander Löser. He holds an M.Sc. in Media Informatics from the BHT and focuses on diagnoses predictions and medical reasoning.

Jan’s background is rooted in software engineering, which provided the technical foundation for his transition into data science. This shift began during his Master’s thesis at Charité Nephrology, where he developed multimodal ensemble models to predict kidney graft loss using Electronic Health Records. Following this, he worked as a Data Scientist at Curalie GmbH, collaborating in interdisciplinary teams to build search algorithms and trust score models for diagnostic prediction. 

Currently, Jan is a Research Assistant and PhD candidate within the DATEXIS research group at the Berlin University of Applied Sciences and Technology (BHT). He is also a member of the Berlin Initiative for Applied Foundation Model Research. His research focuses on the intersection of Natural Language Processing and medicine, specifically creating synthetic medical pathways that simulate the differential diagnosis process. A central aim of his work is to design verifiable rewards to rigorously evaluate medical reasoning in Large Language Models. His recent work includes CliniBench (accepted at EACL 2026), a benchmark for evaluating discharge diagnosis prediction.