PhD Position: Representation Learning & Next-Generation Medical Foundation Models (E13, 100%, 3+ Years)

For our AI-Assisted Healthcare Lab at TUM School of Health and Medicine, we are seeking an outstanding PhD student to develop next-generation multimodal foundation models for digital patient twins in oncology and cardiovascular medicine. The position is part of the EU Horizon Europe project TWIN-X, Digital Twins with Generative AI for Explainable Precision Medicine, a consortium of 18 partners from 12 European countries. Full-time, TV-L E13, fixed-term for 48 months.

PhD Position in Foundation Models and Digital Patient Twins for Precision Medicine

The Department of Diagnostic and Interventional Radiology at TUM University Hospital is recruiting a full-time PhD student (f/m/d) for the EU Horizon Europe project TWIN-X: Digital Twins with Generative AI for Explainable Precision Medicine.

Why this position is unique

This PhD position offers access to large-scale multimodal clinical data, high-end GPU resources, and integration into TWIN-X, an EU Horizon Europe consortium with 18 partners from 12 European countries.

You will work with data from TUM University Hospital and European partner institutions, including:

• Radiological imaging
• Digital pathology
• Genomics
• Laboratory values
• Clinical notes and reports
• Longitudinal patient trajectories from several thousand patients

The project has access to two new NVIDIA B300 servers, additional H100 and H200 GPU servers, and large-scale storage infrastructure. Compute capacity is continuously expanded to minimize bottlenecks.

The position includes funding for conference travel, collaboration with leading European research partners, and opportunities for short research stays at TWIN-X institutions in Greece, Italy, France, the Netherlands, Bulgaria, or Switzerland.

Research vision

The goal is to develop foundation-model architectures for digital patient twins in oncology and cardiovascular medicine. These models should learn patient representations across data types, organs, diseases, and time.

The models should capture:

• Disease trajectories and prior medical history
• Uncertainty and missing information
• Signals relevant to diagnosis and prognosis
• Potential treatment response

The project may include patient-level representations derived from imaging, pathology, genomics, laboratory values, clinical reports, and longitudinal events.

Potential methodological directions include:

• Architectures for heterogeneous and asynchronous clinical data
• Cross-attention models
• Mixture-of-experts systems
• Temporal transformers
• JEPA-style architectures
• Self-supervised and contrastive learning
• Masked-modelling and generative pretraining objectives

Candidates are strongly encouraged to contribute and pursue their own research ideas.

Your responsibilities

• Develop deep learning methods for multimodal and longitudinal patient modelling
• Build and evaluate foundation models on large-scale clinical datasets
• Work with radiology, pathology, genomics, laboratory, and clinical text data
• Design clinically meaningful benchmarks and robust evaluation frameworks
• Publish at leading machine learning and medical AI venues
• Collaborate with clinicians, computer scientists, and European research partners
• Contribute to TWIN-X deliverables and present research internationally

Your profile

We seek a candidate with exceptional analytical ability, excellent academic performance, and a strong technical background.

• Master’s degree with excellent grades in computer science, mathematics, physics, engineering, medical informatics, biomedical engineering, or a related field
• Very strong undergraduate and graduate academic record, particularly in quantitative subjects
• Strong Python skills
• Experience with deep learning frameworks, preferably PyTorch
• Solid foundations in machine learning, statistics, linear algebra, and model evaluation
• Interest in foundation models, representation learning, multimodal learning, generative AI, or longitudinal modelling
• Ability to work independently and learn complex methods quickly
• Excellent English-language skills

German-language skills and prior experience in medical AI are helpful but not required.

We offer

• Full-time position according to TV-L E13 for 48 months
• Opportunity to complete a PhD at TUM University Hospital and the Technical University of Munich
• Access to large-scale multimodal clinical datasets
• Access to high-end GPU and storage infrastructure
• Integration into the TWIN-X consortium with 18 partners from 12 countries
• Opportunities for short research stays at partner institutions
• Funding for international conferences and workshops
• Flexible working hours and options for remote work
• Close collaboration with clinicians, AI researchers, and European partners

Supervision

The position is embedded in the medical AI research environment of TUM University Hospital and the Department of Diagnostic and Interventional Radiology.

Prof. Dr. Lisa Adams
Professor of Radiology
Deputy Director of Radiology, TUM University Hospital

PD Dr. med. Keno Bressem
Radiologist and Coordinator of the TWIN-X project
TUM University Hospital

Dr. rer. nat. Cosmin I. Bercea
Senior Researcher in Generative AI and Medical Imaging
TUM University Hospital

Position details

Position: PhD Student, f/m/d
Topic: Foundation Models and Digital Patient Twins for Precision Medicine
Project: TWIN-X: Digital Twins with Generative AI for Explainable Precision Medicine
Institution: Department of Diagnostic and Interventional Radiology, TUM University Hospital, Klinikum rechts der Isar
Employment: Full-time
Salary: TV-L E13
Duration: 48 months
Location: Munich, Germany

Application

Please send your application by email to keno.bressem@tum.de or lisa.adams@tum.de

Please submit the following documents:

• Cover letter
• Curriculum vitae
• Complete Bachelor’s and Master’s transcripts
• Degree certificates
• Publication list, if available
• Code portfolio or GitHub profile, if available
• Names and contact details of academic references, if available

Please include all undergraduate and graduate transcripts. Applications without complete transcripts cannot be fully assessed.

Die Stelle ist für die Besetzung mit schwerbehinderten Menschen geeignet. Schwerbehinderte Bewerberinnen und Bewerber werden bei ansonsten im wesentlichen gleicher Eignung, Befähigung und fachlicher Leistung bevorzugt eingestellt.