Bhatt-Duchene Group: Computational and Mathematical Global Health

We employ contrastive learning to study images in both satellites and medical cases. We also make use of behavior from video for early detection of infections, using a range of deep learning techniques. We explore the evolution protein structures with the latest deep learning tools, among a broad range of upcoming data types relevant to public health, human history, and genomics.

The genomic data deluge requires new approaches to disentangle the dominant historical and biological signals in genomic data. We are working on highly computationally efficient approaches of exploring historical signals in genomes, as well as the drivers of genomic novelty in pathogens and their animal hosts.

We develop robust, nonparametric methods within a causal inference framework to evaluate the impact of new clinical treatment rules using observational data. By integrating machine learning techniques, we enhance treatment decision guidance while preserving a causal interpretation.