I’m currently working on optimizing a PySpark job that involves a couple of aggregations across large d