项目介绍

2026年6月4日

About the position

The Research Group AI-Assisted Healthcare at TUM University Hospital (Klinikum rechts der Isar) is seeking a highly motivated PhD candidate to join a BMFTR-funded collaborative research project on transparent, agent-based AI methods to support multidisciplinary tumor boards in oncology.

You will develop methods to transform heterogeneous oncology documentation into structured, time-aligned patient trajectories, combining large language models, semantic harmonization, and transformer-based temporal modeling. You will also contribute to the simulation-based evaluation of the resulting decision-support outputs in realistic tumor board scenarios. The work is carried out in close collaboration with partners at TU Munich and RWTH Aachen.

This is a genuinely interdisciplinary role at the interface of machine learning and clinical oncology, with access to a large multimodal research dataset, substantial GPU resources, and a collaborative scientific environment.

Your tasks

• Design and implement LLM-based pipelines for extracting oncological events (diagnoses, staging, treatments, disease course) from clinical free text
• Develop semantic harmonization strategies using standardized vocabularies (SNOMED CT, ICD-10, LOINC) and cross-institutional mapping
• Build transformer-based temporal models that represent patient trajectories and treatment response
• Develop interpretability and attribution methods linking model outputs to their source documents
• Contribute to the design and analysis of simulation-based reader evaluations together with clinical partners
• Publish results at leading venues and present at international conferences

Your profile

• Completed Master’s degree (or equivalent) in Computer Science, Data Science, Computational Linguistics, Mathematics, or a related field
• Strong programming skills in Python and experience with modern deep learning frameworks (e.g. PyTorch)
• Solid background in machine learning, ideally with experience in NLP, large language models, or sequence modeling
• Interest in clinical and biomedical applications; prior exposure to medical data is welcome but not required
• Very good command of English; German is an asset but not a requirement
• Independent, rigorous, and collaborative working style

Our offer

• A PhD position in a well-funded, nationally networked research consortium with strong clinical and methodological partners
• Access to large-scale multimodal oncology datasets and substantial computing infrastructure
• Structured doctoral training and supervision within the TUM environment
• Full-time salary according to TV-L E13, in line with standard public-sector conditions
• A diverse, interdisciplinary, and supportive team
• The position is initially limited to 36 months in line with the project duration

Contact and application

Please send your application as a single PDF, including a cover letter, CV and transcripts to: s.ziegelmayer@tum.de

Applications will be reviewed on a rolling basis until the position is filled. The selection process is organized in two stages.