TU Munich
academicThe Technical University of Munich (TUM) anchors one of Europe's densest AI clusters, tightly braided with Helmholtz Munich and the Munich Center for Machine Learning (MCML). The single-cell and computational-biology line runs through Fabian Theis's theislab, which is dual-affiliated with Helmholtz Munich's Institute of Computational Biology — most of that output is co-owned, and we flag it joint→Helmholtz Munich in the entries below rather than splitting the group.
Flagship work: Theis's scientific foundation models for cells and proteins — Nicheformer (a single-cell + spatial FM pretrained on 110M cells, Nature Methods 2025), OneProt, scTab, CellFlow and the moscot optimal-transport framework; Matthias Nießner's 3D-generative and reconstruction stack (MeshGPT, GaussianAvatars, ScanNet++); Daniel Cremers's computer-vision group; and Stephan Günnemann's reliable/robust-ML and LLM-safety agenda (Soft Prompt Threats).
Industry gravity is heavy: Nießner is on leave founding SpAItial ($13M seed, 2025, "spatial foundation models") after co-founding Synthesia, and TUM appears as a peripheral co-author on the DataComp-LM data benchmark led by UW/Apple/TTIC.
Your notes
People
- Fabian J. Theis Google Scholar — Director, Helmholtz Munich ICB; Chair of Mathematical Modeling of Biological Systems, TUM; MCML
- Daniel Cremers Google Scholar — Chair of Computer Vision & AI, TUM; Director, MCML
- Matthias Nießner Google Scholar — Professor, Visual Computing & AI Lab, TUM (formerly On leave founding SpAItial (2025, $13M seed); co-founder, Synthesia)
- Stephan Günnemann Google Scholar — Professor, Data Analytics & ML (DAML), TUM; Director, MCML
- Zeynep Akata Google Scholar — Liesel-Beckmann Distinguished Professor (Interpretable & Reliable ML), TUM; Director, Institute for Explainable ML, Helmholtz Munich (formerly University of Tübingen (until 2024))