During my Ph.D. in physics, I studied the origin and evolution of the Universe using deep probabilistic modeling. Uncertainty quantification is key in astrophysics, where the telescope optics and the Earth’s atmosphere complicate signals coming from distant galaxies. But complexity riddles protein structure as well. When the COVID-19 pandemic hit, I was motivated to transfer the probabilistic paradigm to the socially impactful field of computational protein design.
As part of the Frontier Research team at Prescient Design, I develop analysis techniques that combine the efficiency of deep neural networks with principled uncertainty quantification by way of hierarchical Bayesian modeling.