I’m am postdoctoral researcher at Stanford University in the group of Prof. Jure Leskovec. Previously, I completed my PhD at ETH Zurich in the Machine Learning and Computational Biology Lab , and prior to that I obtained an MD at the University of Basel.
Here, you can find my current academic CV.
During my PhD, I developed machine learning models for medical time series and developed methods for representation learning, spiked with topological data analysis. Since joining Stanford CS for my postdoc, I have been exploring novel frontiers of medical AI involving zero-shot and few-shot learning, knowledge injection, multimodal medical reasoning, and large pre-trained models. My current research focuses on developing novel machine learning (ML) models for multimodal medical data, such as time series, graphs, images, and text.
Methodologically, I am interested in all kinds of differentiable models (e.g. Gaussian process adapters, temporal convolutional networks, recurrent architectures, transformers, autoencoders etc.). Specifically, I am curious about their ability to learn meaningful representations, and how they can produce generalizable and transferable predictions that are uncertainty-aware. Besides, I am also exploring more mathy flavours of ML using topological data analysis or path signatures.
This is what wordle thinks about my current research in terms of paper titles: