Ephemeral Dataset, I like that! Is this like splitting a dataset for example, then training/testing, when done deleting. Making sure the entire pipeline is reproducible, but without storing the data long term?