Architecting Secure and Scalable AI Chatbot for Diversity Data Using Fine Tuned LLMs With RAG Framework and Lambda Functions

Abstract

In this paper, the authors present a web-based AI chatbot architecture development to understand the challenges of building an intuitive virtual agent capable of extracting insights from diversity data. The developed chatbot combines an information retrieval system with a fine-tuned Large Language Models (LLMs) that is specialised in logic calculations to generate responses. This approach also highlights the importance of balancing diversity’s data security with the users’ desire for a conversational experience. Unlike modern open-domain AI agents that interpret the data and generate responses with their own agency, the developed chatbot provides more structured guidance in conversations, addressing security and privacy challenges associated with processing personally identifiable information, as well as cost and performance issues.