Work

Operationalizing Generative AI Architectures: LLMOps, Retrieval-Augmented Systems, and Real-Time Enterprise Integration

Year: 2024
Type: article
Abstract: Generative artificial intelligence moved from demonstration to production during 2023, and the architectural question shifted from how to train models to how to operate systems built around them. This paper presents a reference architecture for enterprise grade generative AI that combines retrieval augmented generation (RAG), an operational discipline for large language models (LLMOps), and real time integration with enterprise platforms. We describe the retrieval augmented core, comprising an indexing pipeline, a vector store, a retriever and a generator, and explain how grounding generation in enterprise data addresses factual reliability and knowledge freshness without retraining. We then define an LLMOps layer covering deployment, evaluation, monitoring, versioning and governance, and situate the system within an API first and event driven integration layer connecting to customer relationship management (CRM), enterprise resource planning (ERP) and telecom operations and business support systems (OSS/BSS). The paper analyses scalability, security and cost optimisation as first class architectural concerns, including caching, prompt routing, retrieval quality and guardrails. A qualitative evaluation contrasts prompt only, fine tuning and retrieval augmented strategies across cost, freshness, traceability and operational effort. The analysis indicates that retrieval augmentation, combined with a disciplined LLMOps layer and event driven integration, offers the most controllable route to dependable enterprise generative AI, while introducing new responsibilities around retrieval quality, evaluation and data governance
Author: Hemant Soni
Language: English
Open Access status: bronze
OpenAlex ID: W7168024558
FWCI: 0.32
Cites: 14
Cited by: 4
Related to: 10