| Abstract Scope |
Density functional theory (DFT) is a powerful tool for predicting materials properties, but running large numbers of simulations requires significant manual effort for preparing inputs, managing calculations, checking errors, and analyzing results. These tasks often limit the efficiency of high-throughput computational materials research. In this work, we develop an autonomous multi-agent framework for DFT simulations that automates the complete computational workflow. A central VASP agent coordinates specialized sub-agents responsible for job preparation, file management, property evaluation, simulation monitoring, and interaction with an HPC cluster. The structured DFT database is further used to develop ML models that learn relationships between composition, electronic structure, and materials properties enabling AI agent to recommend new pertinent calculations enabling sharpened materials search. By combining autonomous workflow management, intelligent error handling, data curation, and machine learning–guided simulation planning, this framework provides an efficient approach for large-scale computational materials research. |