EMOS: A Modular Framework for Interoperable AI Workflows in Electronic Materials Discovery

Artificial intelligence (AI)-assisted electronic-materials discovery relies on databases, generative models, property predictors, and device simulators that are often difficult to combine. We introduce the Electronic Materials Ontology System (EMOS), an open-source framework designed specifically to make these tools interoperable for electronic-materials discovery through standardized information-unit contracts, a shared property dictionary, and reusable workflow features. In its current version, the framework integrates six inorganic databases, seven generative models, and five property predictors, and is available through a web interface, node editor, and backend application programming interfaces (APIs). We demonstrate both general discovery workflows and an end-to-end materials-to-device screen for metal-oxide-semiconductor field-effect transistor channel materials. The latter composes retrieval, property prediction, stability filtering, synthesisability assessment, and device simulation to reduce 783 structures to two device-evaluated candidates. Finally, we outline how experiment-aware information units could extend this architecture toward reproducible, application-led closed-loop discovery. It provides practical digital workflow composition today while establishing a community-extensible path toward integration with experimental materials research.