This study describes how a machine learning–driven technical case study led to the discovery of a new class of thermoelectric materials. Researchers developed an accurate predictive model to estimate thermoelectric properties and rapidly assess millions of potential materials, helping de-risk exploration of new chemistries. The model successfully identified two novel thermoelectric materials that were later synthesized and experimentally validated. Made openly available online, the tool has been widely reused across thousands of projects, with more than 2,500 researchers generating over 10,000 predictions. By combining rapid assessment, open access, and scalable reuse, the project demonstrates how machine learning can uncover non-obvious materials and accelerate discovery across the resear

Join for free to read