Using Domain Specific Languages to Enable Artificial Intelligence for Polymer and Catalyst Design

Abstract

Traditional research workflows in polymer chemistry, which often rely heavily on trial-and-error, stand to be profoundly reshaped through continued advances in automated experimentation and artificial intelligence (AI). However, the seamless integration of these advances within daily research activities still faces significant hurdles. To overcome these obstacles, we have focused our efforts in key areas which would be impactful for experimentalists. First, we have developed new systems and software for facilitating automated synthesis using readily available lab equipment, both for polymerization in continuous-flow reactors as well as small molecules in flow to batch systems. Second, we show how the use of a domain-specific language (DSL) can solve a host of issues surrounding experimental data representation and management—facilitating more straightforward development of AI models. Finally, we demonstrate how experimental data represented in a DSL can be utilized to develop effective generative models for materials and catalyst design.