In this process, sampled structures through the previous steps are internally stored using an efficient data structure. These stored structures are then refined using high-level DFT calculations in a way compatible with established databases to generate the final dataset.