At the University of Regensburg’s White-Box AI Group, I extended IGANN with expert-defined constraints through convex optimization. I also built a study platform and evaluated whether non-experts could steer the model toward domain rules in a controlled user study.
At Ertebatrasa, I digitized paper records and built a Python recommendation-and-pricing tool that the sales team adopted and continued using. I trained and led junior engineers as the company grew.
My projects include an MLOps service for deploying my interpretable thesis model on Azure AKS and an LLM pipeline that extracts source-traceable requirements from procurement PDFs. My work spans research and applied data products in telecom and medical domains.
