At DATEXIS / Berliner Hochschule für Technik, I develop and evaluate a healthcare AI pipeline that transforms doctor–patient dialogue into structured clinical documentation and downstream ICD predictions.
I build Python workflows for clinical text processing, information extraction, model inference, and evaluation. I assess generated documentation with ROUGE, BERTScore, medical concept-based metrics, and systematic error analysis.
I investigate how information is preserved across dialogue, LLM-generated notes, and ICD prediction, including omission, hallucination, and information drift. I also analyze explainability to understand how clinical information contributes to predictions.
Previously, at B¨orm Bruckmeier, I developed and maintained healthcare and medical information software, including application functionality and database workflows. At CodeBlaQ Technologies, I cleaned and analyzed datasets for exploratory analysis and data-driven insights.
