About
I’m an applied AI researcher and data scientist building trustworthy and efficient medical language models. I’m a PhD candidate at the Chair of AI in Medicine at the Technical University of Munich (TUM), supervised by Prof. Daniel Rueckert, and a Research Scientist at TUM University Hospital. There, I turn research into working systems, such as a customer-facing medical report generation system with automated billing that I developed hand in hand with physicians.
My research asks what it takes to make LLMs reliable enough for medicine. I’ve published on fine-tuning with limited data (TACL) and on unintended memorization of sensitive information (EACL 2026). My current projects include aligning language models through differentiable self-discrimination, quantifying the language barrier in LLM knowledge integration, and building medical knowledge graphs for grounded predictions. In cooperation with Springer Medicine, I’m also working on making agentic clinical decision support traceable and verifiable.
Before medical NLP, I built hybrid physics–machine learning models and developed driver-assistance safety software at Bosch. I studied Mechatronics at BME Budapest and Robotics, Cognition, Intelligence at TUM. I’m a Software Campus Fellow and co-authored the proposal for CLARITY, a €2M BMFTR-funded project on integrity and traceability in multimodal clinical records, starting in 2027.
Research interests
- Trustworthy and efficient medical LLMs
- Fine-tuning with limited data
- Alignment of language models
- Agentic clinical decision support
- Observability and verifiability of LLM agents
- Medical knowledge graphs
- Memorization and privacy in language models
- Multilingual knowledge in LLMs
Latest news
-
award
I was recognized as a Top Reviewer for NeurIPS 2026, which came with a complimentary conference registration. Looking forward to the talks and papers this December!
-
grant
Our project proposal CLARITY (AI-based Contextual Artifact Reduction for Integrity and Traceability in Multimodal Clinical Records), which I co-authored, was funded with €2 million through the BMFTR Zukunft eHealth program (2027–2030).
-
workshop
I attended the GPU Programming Workshop by LRZ and NHR@FAU, covering accelerated computing with OpenACC, OpenMP, Kokkos and CUDA C++ on NVIDIA GPUs. I really enjoyed learning more about these topics and getting hands-on experience on a GPU cluster.
-
grant
I was selected as a Software Campus Fellow, receiving a BMFTR Tech Leadership Grant of €115k for 2026–2028.
-
publication
Our survey and practical guide Fine-tuning Large Language Models with Limited Data is published in TACL.
Read more
Selected publications
-
TACL
Fine-tuning Large Language Models with Limited Data: A Survey and Practical Guide
Transactions of the Association for Computational Linguistics, 14, 341–377, 2026
-
EACL '26
Unintended Memorization of Sensitive Information in Fine-Tuned Language Models
Proceedings of the 19th Conference of the European Chapter of the ACL (Volume 1, Long Papers), 6461–6480, Rabat, Morocco, 2026