This document summarizes the work of Zachary Ulissi and his research group on developing computational workflows and machine learning methods to accelerate catalyst discovery. Key points include: - The group is generating large datasets of catalyst descriptors like adsorption energies to train graph neural network models that can predict properties and screen for new catalysts. - Their GASpy software coordinates density functional theory calculations using dependency graphs to efficiently generate data. - Preliminary graph convolution models show accuracy comparable to DFT for predicting adsorption energies and surface energies. - The group is applying these methods to problems like oxygen evolution reaction catalyst screening to potentially find new active compositions beyond what is searched manually.