Professor Alexander Kvashnin from the Energy Transition Center and PhD student Faridun Jalolov have introduced a novel machine learning method for studying the properties of complex materials — polycrystals, composites, and multiphase systems — achieving accuracy comparable to quantum-mechanical methods but applicable to much larger systems. Unlike traditional methods limited to a few hundred atoms, this approach uses moment tensor potentials to perform active learning on local atomic environments, allowing for the identification and correction of inaccuracies in real time. The research was covered by Naked Science (RUS), CNews (RUS), Swift Telecast (ENG), and Carbon Chemist (ENG).