My M.Sc. thesis at Dokuz Eylül University is a host-response infection classifier that reached ≈0.95 external macro-AUC across 8 independent cohorts. I integrated public cohorts and trained a leakage-free ensemble designed to classify a single patient sample without batch correction.
Earlier work on the infection classifier was published at IEEE ICHORA 2026. I also brought its probability calibration error to 0.037 and added an abstention rule for low-confidence cases.
At MediaFG, where I worked as a Full Stack Developer from 2024 to 2026, I built and maintained client applications with Django and Next.js. I also designed and built the agency’s internal management system as its only developer, replacing spreadsheets and chat threads with a production workflow.
My other work includes a Llama-based RAG pipeline and a race-photo finder using face embeddings and the Google Drive API. I’ve also taken freelance ML and data projects from raw data through deployment, and supported model deployment and retraining workflows as an MLOps Intern at Kuartis.

