AI-Powered Full-Stack Website Generator Using MERN Stack

Abstract

Recent advancements in multimodal artificial intelligence, large language models (LLMs), and real-time code generation have transformed the landscape of automated soft- ware development. Traditional website creation involves significant manual work from designers, developers, and front- end engineers, leading to high costs and prolonged delivery timelines. Existing systems for UI-to-code generation and natural language based website synthesis have made progress but remain constrained by model latency, limited flexibility, and in ability to self-evolve. This paper introduces WebAI, a novel SaaS platform built using the MERN stack and powered by low- latency Groq LPU-driven LLM inference, enabling real-time generation of complete, production-ready websites from either natural-language descriptions or UI images. A key innovation of WebAI is its recursive website generation engine, allowing each generated website to contain an embedded generator capable of creating further websites. The system integrates multimodal LLM reasoning, computer vision (CV)-based UI understanding, abstract syntax tree (AST)-driven code validation, scalable cloud architecture, and a template-driven synthesis module. This paper presents an in-depth analysis of WebAI’s design, architecture, methodology, performance evaluation, datasets, and contributions to the emerging field of autonomous web engineering.