Models Are Data Too: Towards Expressive Querying of Machine-Learning Models

Machine learning generates large amounts of models that should be managed by sound data management practices. In particular, we should be able to query these models using expressive, declarative query languages. In this paper, which is based on a keynote we delivered at PODS 2026, we introduce this new kind of database querying, mainly by surveying existing ideas and pointing at research challenges. We also briefly foray into the querying of program code and query expressions.