The prototypical discovery workflow for small molecules involves iterating through design-build-test loops. AI has a role to play in (a) property prediction, (b) molecular design, and (c) synthesis planning, among others. This talk will summarize some of our work to expand the role of AI in augmenting each of these tasks, both individually and simultaneously. I will describe how predictive models and formal optimization frameworks can help make more informed, quantitative decisions in iterative design cycles that balance properties of compounds against their synthetic cost.
