An apparatus includes a database and a processor. The database stores a set of services, each of which produces outputs when provided inputs. The processor determines a machine learning policy that can be applied to a first service to determine a descriptor file assigned to the service. The first service can produce a first output when provided a first input. The descriptor file includes the first output type and the first input type. The processor applies the machine learning policy to the set of services to determine a set of descriptor files. The processor receives a request from a user for a second service that produces a second output given a second input. The processor then locates a second descriptor file that includes the second output type and the second input type, determines a third service assigned to the descriptor file, and sends the third service to the user.