REACH OUT FOR THE STARS

Hello, I'm Niklas Bubeck

PhD Candidate · AI for Medical Imaging · TUM

Researcher at the Technical University of Munich working on generative models, cardiac MRI reconstruction, and tokenization for medical imaging.

View Research Get In Touch

Scroll to explore

About Me

Passionate researcher dedicated to advancing scientific knowledge

I am a dedicated researcher with expertise in the field of AI in Medicine and Healthcare. My work focuses on medical imaging and the implications of generative AI. I have multiple years of experience the academic and industrial world and am working with Daniel Rueckert and Jiazhen Pan.

Education

Now: Ph.D. in AI in Medicine from TUM

2020 - 2023: M.Sc. in Robotics, Cognition, Intelligence from TUM

2016 - 2020: B.Sc. in Medical Engineering from FAU Erlangen

Experience

Now: PhD Researcher at the TUM Chair for AI in Healthcare and Medicine / MCML

Since 2026: AI Consultant, Freelance

2021 - 2022: Software Engineer, ADAS HIL Test Automation at IAV AG

2019 - 2021: Software Engineer, Software & IoT Systems at Siemens AG

2018 - 2019: Teaching Assistant at the Chair of Technical Electronics, FAU

Leadership

2022 - 2023: Head of Industry at TUM.ai — led a team of 12 on projects with Infineon, RTL, Pro7Sat1, QuantCo, Recogni and UniCredit

2018 - 2021: Head of Driverless at High-Octane Motorsports — led 10 engineers building an autonomous racecar for Formula Student

Total Citations

h-index

i10-index

Papers

Top Coauthors

Coauthors from the Technical University of Munich and collaborating institutions.

Research Areas

Generative modeling for medical imaging — and whether the results can be trusted

Cardiac MRI

Recovering full 3D cardiac anatomy from sparse acquisitions: work with NVIDIA interpolates in a diffusion model's latent space, while a parallel line reads cardiac function straight from undersampled k-space.

k-Space
Diffusion Models
3D Reconstruction

Generative AI

Where does reconstruction end and generation begin? My DGM4MICCAI work maps that spectrum for cardiac MRI; related projects grow glioblastoma under physics-guided constraints, so tumour dynamics drive the image.

Image Synthesis
Physics-Guided
Autoregressive Generation

Tokenization

Compressing medical volumes into compact discrete representations generative models can use directly. How far can compression go before anatomical fidelity breaks down? Publication forthcoming.

Autoencoders
Discrete Representations
High Compression

Trustworthy Medical AI

Medical AI that clears a benchmark can still fail patients. My red-teaming agents probe medical language models for what benchmarks miss (Nature Health, 2026); with Harvard Medical School I study how reconstruction shapes fairness.

Red-Teaming
Fairness
Evaluation

Publications

Recent contributions to academic literature

Addressing benchmarking gaps in large language models for health and medicine with dynamic red-teaming Jia-Zhen Pan, Bailiang Jian, Paul Hager, Yundi Zhang, Che Liu, F. Jungmann et al. Nature Health · 2026 · 14 citations
TumorFlow: Physics-Guided Longitudinal MRI Synthesis of Glioblastoma Growth Valentin Biller, Niklas Bubeck, Lucas Zimmer, Ayhan Can Erdur, Sandeep Nagar, A. Meyer-Baese et al. arXiv preprint · 2026 · 3 citations
Evaluating the Impact of Medical Image Reconstruction on Downstream AI Fairness and Performance Matteo Wohlrapp, Niklas Bubeck, Daniel Rueckert, W. Lotter arXiv preprint · 2026
No Image, No Problem: End-to-End Multi-Task Cardiac Analysis from Undersampled k-Space Yundi Zhang, S. Kafali, Niklas Bubeck, D. Rueckert, Jia-Zhen Pan arXiv preprint · 2026
Beyond Benchmarks: Dynamic, Automatic And Systematic Red-Teaming Agents For Trustworthy Medical Language Models Jia-Zhen Pan, Bailiang Jian, Paul Hager, Yundi Zhang, Che Liu, F. Jungmann et al. arXiv preprint · 2025 · 6 citations
Reconstruct or Generate: Exploring the Spectrum of Generative Modeling for Cardiac MRI Niklas Bubeck, Yundi Zhang, Suprosanna Shit, D. Rueckert, Jia-Zhen Pan DGM4MICCAI@MICCAI · 2025 · 3 citations
Latent Interpolation Learning Using Diffusion Models for Cardiac Volume Reconstruction Niklas Bubeck, Suprosanna Shit, Chen Chen, Can Zhao, Peng-Fei Guo, Dong Yang et al. arXiv preprint · 2025 · 3 citations
A Biophysically-Conditioned Generative Framework for 3D Brain Tumor MRI Synthesis Valentin Biller, Lucas Zimmer, Can Erdur, Sandeep Nagar, D. Rückert, Niklas Bubeck et al. BraTS-Lighthouse/AIMS-TBI@MICCAI · 2025

Project sites

Live pages from repositories with GitHub Pages enabled — swipe or use the arrows

Loading projects from GitHub…

Get In Touch

Let's collaborate and advance research together

Email

niklas.bubeck@tum.de

Office

Walther-von-Dyck-Straße 4
Galileo 5. OG
85748 Garching bei München

Phone

People call me Bubi, but you can call me anytime :P