About
I'm a PhD student in Fabian Theis' lab at Helmholtz Munich, supported by the German National Scholarship Foundation. My work sits at the intersection of deep learning and genomics, where I am building models that are not only accurate but interpretable. I previously trained with Roland Eils at the Digital Health Center (BIH & Charité, Berlin), and spent six months in David Baker's lab at the Institute for Protein Design (University of Washington), where I worked on generative protein design.
I'm always happy to chat about research, collaborations, or ideas. Feel free to reach out.
Research Interests
Publications
# denotes equal / co-first contribution
An Interpretable Omnigenic Neural Network Architecture for the Human Genome
medRxiv (2026) · doi: 10.64898/2026.07.28.26359187
The Human Pancreas Cell Atlas Defines a Healthy Reference Framework for Disease Contextualization and Translational Benchmarking
bioRxiv (2026) · doi: 10.64898/2026.06.22.733853
annbatch unlocks terabyte-scale training of biological data in anndata
arXiv (2026) · doi: 10.48550/arXiv.2604.01949
A Predictive Atlas of Disease Onset from Retinal Fundus Photographs
Lancet Digital Health (2026) · doi: 10.1016/j.landig.2025.100962
Biologically Informed Variational Inference Enables Interpretable Cell Phenotyping and Discovery
bioRxiv (2025) · doi: 10.1101/2025.06.10.657924
Multistate and Functional Protein Design Using RoseTTAFold Sequence Space Diffusion
Nature Biotechnology (2024) · doi: 10.1038/s41587-024-02395-w
Software
Contributor
lueckenlab / patpy
Toolbox for single-cell data analysis on sample level.
scverse / annbatch
Minibatch loading for on-disk AnnData files, enabling terabyte-scale out-of-core training directly on standard biological data formats.
theislab / embpy
A unified Python toolkit for generating and comparing biological embeddings across genes, proteins, molecules, morphology, and single cells.
luisherrmann / udm
Unified Data Module for multimodal training on UK Biobank data, synchronizing modalities across data loaders built on PyTorch Lightning.