My goal is to build the foundational machine-learning methods that enable the next generation of medicine—especially for complex diseases where today’s models and clinical pipelines still fail (cancer, autoimmune disease, neurodegeneration, and chronic inflammatory disorders).

Research focus: foundation models for the immune system

A central theme of my work is modeling the human immune system. The immune system is both our most powerful defense and, in many conditions, a major driver of pathology—e.g., severe outcomes in viral disease and the self-damage seen in autoimmunity. Despite breakthroughs like immunotherapy, the immune system remains hard to reason about because it is context-dependent, multi-scale, adaptive, and strongly shaped by interaction networks across cells, tissues, and time.

My bet is that AI can help move beyond intuition-limited reasoning—by integrating large, heterogeneous biological datasets and extracting mechanistic, testable structure from them.

Technical interests

I work on ML methods at the intersection of representation learning, generative modeling, and structured biological data, including:

• Geometric deep learning for molecular and cellular structure (equivariance, symmetry-aware models)

• Diffusion and other generative models for biological structure and distributional uncertainty

• Multimodal foundation models for immune data (single-cell transcriptomics, protein/epitope, perturbations, clinical context)

• Graph-based learning for interaction systems and cell–cell / gene–gene networks

• Robust evaluation under distribution shift (dataset bias, label noise, uncertainty-aware predictions)

Collaboration

This work is inherently interdisciplinary. I’m especially interested in collaborating with:

• Immunologists / experimental labs with perturbation screens, immune profiling, or longitudinal cohorts

• Clinicians / translational groups working on immunotherapy response, autoimmune stratification, or severe infection

• ML researchers interested in geometric models, multimodal learning, and evaluation under shift

If you have data, a well-defined therapeutic question, or a hard modeling bottleneck, I’m happy to talk.

AI safety & biosecurity

In parallel to my biomedical research, I take the risks from advanced AI systems seriously. I’m part of the Munich AI safety community and previously participated in a technical AI-safety reading group for two years during my Master’s in Amsterdam. If you’re looking to enter AI safety research or want pointers toward high-quality communities and institutions (technical or governance-oriented), feel free to reach out—I’m happy to connect you to the right people.