The first is probably done using pipeline controllers, the second using Datasets or HyperDatasets. Its not very clear how the last one is achieved, especially on the searchable data catalogs.