Title:

Teaching Machines Chemistry

Speaker: Juri Smirnov
Date: 2023-10-20

Abstract:

Do language and chemistry have something in common? As it turns out they do, and machines are excellent at picking up the semantics out of the strings encoding molecular structures. Those strings resemble sentences which follow particular grammatical construction rules. We capitalise on the enormous progress in artificial intelligence, with a particular focus on generative models. Our goal is to teach a machine chemical features of molecules and use its ability to generate new structures to find new molecules for our applications. As a concrete example, I will present a search for optimal crystalline molecules for light dark matter detection, as well as our exploratory search for bioactive molecules that we conduct with our industry partners.