My work focuses on “programmatically interpretable learning”: solving learning tasks using models that can be expressed as programs in a high-level domain-specific language.