Tangible Successes from Polymer Informatics

For more than a decade, Citrine Informatics has been creating software tools for machine learning-accelerated materials innovation. Materials informatics applications are characterized by complex constraints, disjoint design spaces, and small, heterogeneous, uncertain data. The intersection of these challenges has required substantially different methodological approaches as compared to traditional artificial intelligence applications. Citrine has brought to market a platform that allows materials practitioners to benefit from these technologies without requiring users to write software or be expert in machine learning techniques.

In this presentation, real world successes will be discussed. Specifically, customer successes with switching to bio-based feedstocks, creating novel cellulosic esters, removing phthalates from PVC, accounting for varied post-consumer recyclates, and leveraging lightweight polymer composites for automotive applications will be presented. Details of both the underpinning information science and statistical learning will be presented in tandem with Citrine's no-code workflows.

Materials informatics is in the midst of massive growth, in terms of an increasing number of providers, increasing levels of funding, and substantive changes in the types and power of models. While "[p]rediction is very difficult, especially about the future", personal perspectives about the short term consequences of this sea change will be discussed.