Generative AI in Action By Amit Bahree #BookSummary
Pankaj Gajjar
Husband|Father|Speaker|Enterprise Architect (TOGAF®)|MDM(PIM/DAM/MXM) Architect|ACE(Multi Cloud)|ex-AWS CB|Principal Solution Architect @Datastax|GenAI |AI Consulting | Distributed DB
Application architecture for generative AI apps
Copilot demonstrates how generative AI architecture can build enterprise applications and solutions. It uses a different application stack that works with copilots to create the new enterprise architecture stack. This stack is for GenAI apps, which use Copilot as a counterpart to the LAMP stack.
The GenAI app stack includes four layers that cooperate to make the application stack function—the model, orchestration, grounding, and response filtering layers.
The orchestration layer is one of the critical and foundational components of the GenAI stack. It handles and organizes different processes, AI services, and platforms to enable a dependable and coherent experience.
The area of orchestration frameworks is new and evolving, with many changes and innovations taking place. Some of the frameworks that are more widely used today are SK, LangChain, and LlamaIndex.
By using plugins than can handle the intricacies of the source systems, their protocols, and other details, the grounding layer facilitates data integrations and preprocessing for RAG deployments in the enterprises. It also oversees the embeddings and the related vector databases.
The model layer offers a platform for using multiple models from various sources—from managed and fine-tuned models to BYOM (bring-your-own-model) for enterprises. These models can all be accessed through strong APIs that guarantee compliance and security.
The response filtering layer ensures quality and moderates content, essential for building confidence in GenAI applications. Furthermore, it involves using classifiers and NLP tools to check the outputs for correctness, fairness, and suitability.
An AI Center of Excellence can help enterprises comprehensively integrate LLMs and GenAI into their applications. By addressing technical, cultural, and ethical challenges, enterprises can use AI to enhance innovation and competitiveness, ensuring lasting success in an increasingly AI-powered world.