Scalable learning systems
for real-world biology.
I am a PhD Researcher & Research Engineer in AI for Science at DKFZ & EMBL Heidelberg. I specialize in bridging the gap between expressive geometric modeling and production-scale systems, translating high-dimensional biological signals into robust, uncertainty-aware infrastructure for clinical and mechanistic discovery.
Research Philosophy
I build learning systems that transform messy, large-scale biological data into reliable, production-grade infrastructure. My focus: designing GNN and Transformer pipelines that scale to 10M+ cells and transcripts while preserving mechanistic interpretability—from spatial omics segmentation to multi-modal integration.
Scientific Dimensions
Spatial & Single-Cell Omics
End-to-end pipelines for spatial transcriptomics, scRNA-seq, and multi-modal integration. I design models that capture cell neighborhood structure and molecular heterogeneity to drive biological discovery.
Scalable Graph Neural Networks
Multi-GPU training and inference for heterogeneous GNNs on graphs with 10M+ nodes and 100M+ edges. Achieved 1000x speedups for spatial phenotyping tasks previously considered computationally intractable.
Open-Source Tooling
Architecting production-grade frameworks (Segger, SageNet, scGCN) and contributing to community standards like SpatialData to ensure reproducibility and extensibility.