Research Interests
I work in the field of Bayesian deep learning. Most of my research combines representation learning and probabilistic modeling. Areas of interest are as follows:

New approaches to approximate Bayesian inference, drawing on stochastic optimization, stochastic gradient Markov chain Monte Carlo, and variational methods; often involving ideas from statistical physics (SGD as approximate inference; iterative amortized inference, QMC variational inference, perturbative variational inference, cold posteriors etc.)
Neural image and video compression based on variational autoencoder models (we have published one of the first neural video codecs and--as of June 2020--the best-performing image codec)
Machine learning for time series and learning under distribution shift (disentangled sequential autoencoders, dynamic word embeddings, dynamic topic models, modeling event data)
Deep generative modeling in the natural sciences, in particular physics, chemistry, and climate science.