Use parallelization -- either in the python or in the shell processes it creates, or both. Our work should be *very* friendly to parallelization at the zip and at the json file level. https://stackabuse.com/parallel-processing-in-python/