DECENTER AI Package for AI Microservice

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.

AI Optimization method

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.