I am a highly motivated PhD student at the University of Basel under the supervision of Volker Roth. Currently, I am working to make machine learning models more expressive and interpretable at the same time.

During my PhD

As a first result we improved sparsity in Bayesian graphical models with Conditional Matrix Flows (see NeurIPS publication). In parallel, together with Fabricio Arend Torres we developed a physics-informed deep learning model that conserves the continuity equation by construction (see ICLR Spotlight - top 5% papers).

We also developed a comprehensive library of (conditional) Normalizing Flows in PyTorch named FlowConductor.

As a side project, I recently helped Human Rights Watch developing a software to detect village burnings in Darfur, Sudan.

Before my PhD

Previously, I worked on my MSc thesis with Gunnar Rätsch (ETH) at the interface between VAEs and meta-learning. As a research assistant I worked with Aurelian Lucchi and Thomas Hofmann (ETH) on a inter-disciplinary project about deep learning for gravitational wave physics. Previously, I also worked with Frank Schweitzer (ETH) to help the group with data analysis tasks in three different projects.