Various methods, apparatuses, and media for implementing a machine-learning model execution module are provided. A processor is configured to generate a machine-learning model. The machine learning model includes data related to a requester's access to one or more ephemeral roles. The processor receives a request from the requester to access the one or more ephemeral roles within the machine-learning model. The processor also determines the requester's group or role membership status within an organization. The processor also dynamically evaluates the received request with the machine learning model in real time to grant or deny access to the one or more ephemeral roles based on the membership status of the requester.