MLMD: a programming-free AI platform to predict and design materials

Accelerating the discovery of advanced materials is crucial across industries, yet only a small portion of the chemical space is experimentally explored due to time-consuming, labor-intensive R&D. While many AI toolkits exist, they often focus narrowly on property prediction, require programming expertise, and struggle with limited data. We develop MLMD, a programming-free AI platform for materials design that supports end-to-end discovery of novel materials with high-potential properties via model inference, surrogate optimization, and active learning for data-scarce scenarios. MLMD integrates data analysis, descriptor forecasting, automated hyperparameter optimization, and property prediction, offering a web-based, code-free interface. It facilitates integration of experiments/computation with inverse design for single or multiple objectives and demonstrates effectiveness across diverse materials (perovskites, steel, high-entropy alloys). MLMD aims to be an essential tool advancing materials informatics.

Jiaxuan Ma, Bin Cao, Shuya Dong, Yuan Tian, Menghuan Wang, Jie Xiong, Sheng Sun