Interfaces and APIs are needs to be implemented to make an AI microservice from a trained AI model. There are few platforms which supports providing an AI model, however to build an AI (micro)service it needs to include data processing before and after the model processing. DECENTER provides a Python package which enables packaging AI microservice from an AI method. It provides network APIs to access embedded AI method, model management, and pre-/post-data processing.
DECENTER’s goal on AI includes bringing the intelligence near to the device or to the service endpoint. There are several aspects needs to be considered when bringing AI close to edge including prompt response between users and service and privacy concerns. SNU is investigating on suitable AI optimization method for DECENTER such as splitting AI, model compression and acceleration.