Described herein are systems and methods for profiling structured or semi-structured datasets. An example computer-implemented method includes grouping, using a machine learning classifier, a plurality of tables in a dataset that are associated with an object into a cluster, where each of the tables of the cluster includes respective data and respective metadata, the respective metadata including at least one respective attribute, generating a metadata-profile for the cluster, where the metadata-profile includes the at least one respective attribute of each of the tables of the cluster; and querying the cluster using the metadata-profile.