Researcher | Founder
My dream is to use my background in mathematics and software development to design sophisticated tools that can improve data-driven healthcare and help facilitate the urgent action needed to combat climate change.
I'm completing my PhD at the Technical University of Munich, researching topological and geometric deep learning as a member of the AIDOS Lab at the University of Fribourg. I'm also a co-founder of Krv Labs, where we accelerate patient discovery in clinical trials by transforming complex eligibility criteria into reviewable longitudinal EHR searches.
Always looking for new opportunities to collaborate and exchange ideas—feel free to reach out!
We combine graph methods and topological analysis to cluster the US coal fleet into retirement archetypes. The resulting vulnerability scores identify group-specific levers, helping policymakers prioritize targeted pathways for faster coal phase-out.
We introduce RINGS, a mode-perturbation framework to assess graph-learning dataset quality. Using performance separability and mode complementarity metrics, we provide principled tools for evaluating how well datasets benchmark graph-learning methods.
We use Forman-Ricci and Ollivier-Ricci curvature to analyze physician referral networks and detect regional variation in care delivery. This work introduces 'apparent', an open-source pipeline linking network geometry to local demographics and outcomes.
We present PRESTO, a framework using persistent homology to map the multiverse of latent representations. It enables sensitivity analysis, anomaly detection, and efficient hyperparameter navigation by characterizing embedding variability across ML models.
A topological descriptor for graph distributions using curvature filtrations. Our method enables scalable statistical comparisons and hypothesis testing for evaluating graph generative models, with applications to drug development.
Rust-backed Python library for topological data analysis implementing the Thema pipeline: imputation, scaling, PCA, Ball Mapper, and Cosmic Graph, with optional MCP tooling for agent-driven analysis.
Structural code-quality engine for coding agents that evaluates modules against priorities like self-contained and composable, then guides iterative refactoring using measurable lattice-based verdicts.
Representation-guided imputation library for missing tabular data that generates multiple imputations, computes Euler Characteristic Transform descriptors, and selects the most representative candidate.
Topological representation learning library for EHR and tabular data built on differentiable Euler Characteristic Transform, with NumPy, pandas, and polars workflows.
I studied mathematics and astrophysics at UC Berkeley, then completed a master's in computational and data sciences at Chapman University. Along the way, I worked as an ML consultant at Madiba and as a research developer at Encryptek.
That path led to CHOC, where I worked with Louis Ehwerhemuepha as a grant-funded research scientist on predictive models for pediatric sepsis—and found my way into computational topology. I then joined Bastian Rieck's AIDOS Lab at the University of Fribourg to pursue a PhD in topological and geometric deep learning.
During my PhD, Stuart Wayland, Sidney Gathrid, and I began collaborating on various projects. In one particularly fun project, we used Riemannian curvature and topological data analysis to study the phase-out of US coal power plants (now published in Nature Energy). We decided we loved working together and wanted to keep doing it, so we founded Krv Labs. Krv (Kepler Riemann von Neumann) pays homage to some of the scientists we look up to who refused to accept 'good enough'. We bring that same standard to our work in healthcare, translating methods from mathematics, physics, and machine learning into safe, effective, and evidence-backed AI.