AI-Integrated Jewelry E-Commerce Platform: A Full-Stack Web Application with Generative Design, Conversational AI, and Real-Time Market Intelligence

Overview

The Indian jewelry retail sector, valued at over USD 85 billion, remains characterized by conventional storefronts with limited digital integration. This work presents a full-stack web application developed for Amar Jewellers, a heritage retailer in Sangli, Maharashtra, designed to address market fragmentation through AI-integrated capabilities. The platform implements a three-tier architecture comprising a React-TypeScript single-page application frontend, an Express.js RESTful API backend, and MongoDB persistence layer. The system introduces three primary innovations: an AI-powered Design Studio utilizing the Gemma 3 large language model for custom jewelry visualization generation, a conversational chatbot providing context-aware assistance with live commodity rate integration, and real-time precious metal price feeds via streaming API infrastructure.

Methods and approach

The architecture employs modern web development frameworks with separation of concerns across presentation, application, and data tiers. The Design Studio component integrates the Gemma 3 language model through the Bytez inference API, accepting parameterized input specifications including metal type, style, weight, and budget constraints to generate custom visualizations. The conversational interface, named Amar, incorporates live commodity rate data sourced from the Amar Bullion broadcast streaming API to provide dynamic responses. Authentication mechanisms employ JWT-based multi-role access control distinguishing customer and administrator privileges. The repair tracking module integrates with Google Sheets infrastructure and assigns unique ticket identifiers for lifecycle management. The user interface layer implements Framer Motion-driven animations, Lenis smooth scrolling, and glassmorphism design patterns. Performance evaluation assesses API response latencies for rate retrieval, design generation execution timeframes, and end-to-end repair submission and status tracking workflows.

Key Findings

Experimental evaluation demonstrates sub-second API response latencies for real-time commodity rate fetching operations. AI-powered design generation completes within acceptable timeframes from user-specified parameters. The repair tracking system exhibits full functional capability across the complete lifecycle from initial submission through status tracking and resolution. The integrated conversational chatbot successfully provides context-aware responses incorporating live market data and platform navigation assistance. The system maintains data persistence and retrieval efficiency across the MongoDB persistence layer while supporting concurrent user interactions through the Express.js backend infrastructure.

Implications

The platform demonstrates technical feasibility for integrating generative AI and real-time market intelligence into heritage jewelry retail operations. By implementing AI-driven design democratization and transparent pricing mechanisms, the system addresses longstanding limitations in the traditional jewelry retail sector, particularly for small and medium-sized enterprises lacking independent technology infrastructure. The three-tier architecture provides scalability pathways for expansion beyond single-retailer deployment across broader jewelry enterprise networks. The integration of conversational AI with live commodity pricing establishes precedent for knowledge-augmented retail interfaces within price-volatile commodity sectors. The system bridges contemporary digital retail practices with heritage craftsmanship methodologies, establishing technical infrastructure for heritage businesses to compete in digital-first consumer markets without compromising artisanal production values.