A Hybrid Generative AI and Micro-Frontend Architecture Using Transformer Models for Scalable and Intelligent Retail Web Applications

Abstract

The fast evolution in e-commerce requires web solutions that are intelligent, scalable and personalized enough to cope with the changing consumer behavior. This paper presents a hybrid retail web app development architecture incorporating Generative Artificial Intelligence (GenAI) techniques with Micro-Frontend design principles based on Transformer-based models. The proposed framework utilizes large language models (LLMs) including GPT and BERT variants for real-time product recommendations, intelligent search, dynamic content generation and conversational commerce interfaces. The architecture enhances modularity, fault isolation, and team scalability by breaking down monolithic frontend structures into separate micro-frontends that can be independently deployable. This enables the seamless communication of each micro-frontend in our code base with its corresponding AI inference APIs, allowing personalization that is context-aware at the component level. The system uses federated deployment strategies, edge caching and asynchronous AI pipelines to deliver low-latency performance in high-throughput retail reality. Experimental evaluation shows that we achieve significant improvements in page responsiveness, recommendation quality, and user engagement metrics over monolithic AI-integrated retail platforms. This hybrid architecture specifically combines smart AI capabilities with contemporary frontend engineering practices to deliver a future-proof, robust and maintainable architecture for next-generation retail web end-to-end ecosystems.