ej

Meet ej

ej is a compact decision model for applications that need to choose, classify, or score rather than generate an open-ended response. A developer supplies a record and questions with possible outcomes, then receives probabilities for the options. Several questions about the same record can be answered in one pass.

The model runs on a CPU inside the application's own environment and can work offline after installation. Example uses include routing support tickets, checking a tool request, and estimating urgency. Teams can set thresholds that automate confident outcomes and pass uncertain ones to a person or a larger model. Open weights and a Python library provide the starting point for integration.