Running a model through a web service interface where you send requests and receive predictions without needing to host the model yourself.