From Microservices to Model-Centric Architecture in Scholarly Knowledge and Research Intelligence Systems

Abstract

Scholarly knowledge platforms have evolved into large scale digital ecosystems that support research discovery, analytics, and evaluation across global academic communities. Traditional microservice architectures have enabled modular growth of such platforms, yet they increasingly struggle to accommodate the dynamic, learning driven nature of modern research intelligence workflows. This paper examines the architectural shift from service centric designs toward model centric architectures, where machine learning models, inference pipelines, and knowledge representations become first class system components. Focusing on scholarly knowledge and research intelligence systems, the study proposes a reference architecture, formalizes core inference workflows, and evaluates system level performance across scalability, adaptability, and trust related dimensions. Experimental results demonstrate measurable improvements in latency stability, model evolution velocity, and semantic consistency. The findings highlight how model centric design principles better align system architecture with the epistemic and operational demands of contemporary scholarly ecosystems.