Case Study: Thermo Electric discovers new thermoelectric materials with Citrine Informatics

How Machine Learning Model Made Available to Research Community Discovered New Classes of Thermoelectrics and Helped Thousands of Projects

Citrine Informatics worked with Thermo Electric to address the challenge of discovering new thermoelectric materials. The project aimed to create a machine learning model that could accurately predict material properties to identify novel thermoelectrics that would not be found through conventional intuition alone, and to make this tool openly available to the wider research community.

Citrine Informatics developed a solution by first compiling a dataset of known thermoelectric properties. They then built and trained a machine learning model using domain-knowledge-infused descriptors. The solution successfully predicted, synthesized, and validated two new thermoelectric materials. Furthermore, Citrine made the model available as a public web app, which has since been reused extensively, evidenced by over 2,500 researchers using it to make more than 10,000 predictions, dramatically accelerating materials research.

Thermo Electric