Biography
I’m mainly interested in probabilistic modelling for science and statistical machine learning. As part of my PhD, I’ve shown that most classical dimensionality reduction methods (such as t-SNE, UMAP, MDS and Laplacian embeddings) and modern representation learning algorithms (such as some contrastive methods) have probabilistic interpretations as inference algorithms of probabilistic models akin to Gaussian processes. I have also worked on several case studies on using latent variable models in practice, e.g. single-cell data analysis, climate science, proteins and bioacoustics, and am interested in deployment (e.g. using probabilistic programming).