OPEN FOR NEW OPPORTUNITIESAVAILABLE FOR FREELANCEFRONTEND DEVELOPERREACT & NEXT.JS SPECIALISTOPEN FOR NEW OPPORTUNITIESAVAILABLE FOR FREELANCEFRONTEND DEVELOPERREACT & NEXT.JS SPECIALIST• 
OPEN FOR NEW OPPORTUNITIESAVAILABLE FOR FREELANCEFRONTEND DEVELOPERREACT & NEXT.JS SPECIALISTOPEN FOR NEW OPPORTUNITIESAVAILABLE FOR FREELANCEFRONTEND DEVELOPERREACT & NEXT.JS SPECIALIST• 
Case Study • 02 / EYORIS FASHION
Back to All Work

Eyoris Fashion

A modern luxury fashion e-commerce platform featuring AI visual search, multimodal Gemini catalog tagging, and high-performance full-stack architecture.

Next.js 16React 19TypeScriptNode.jsExpressMongoDBTailwind CSSGoogle Gemini AIClerk AuthFramer Motion
Key Metric Achieved
< 500ms

Real-time AI visual search & structured tagging across 44,000+ catalog items.

RoleFull Stack Developer & AI Engineer
Timeline2024 – 2025
Client / CompanySelf Project / Production Showcase
DeliverablesMonorepo Web App, RESTful API, AI Visual Search Engine, Admin Dashboard
Eyoris Fashion
01 • The Challenge

Building intuitive visual discovery and automated catalog indexing at scale.

Traditional e-commerce search relies heavily on text queries and manually tagged catalog attributes, creating friction when users search by style, visual aesthetics, or uploaded photos.

Building an automated pipeline capable of parsing raw customer uploads, recognizing precise garment categories, extracting dominant color palettes, and returning matching product variants with low latency required careful AI integration and optimized database queries.

02 • The Solution

Multimodal AI Integration with Next.js 16, Google Gemini, and Express/MongoDB.

Engineered a full-stack monorepo featuring a Next.js 16 frontend with dynamic animations and an Express/MongoDB REST backend.

Integrated Google Gemini vision models using strict JSON schema enforcement to perform zero-shot visual categorization and dominant color extraction directly from uploaded images. Combined this with Clerk authentication, variant inventory tracking, and deterministic catalog seeding.

03 • Implementation & Architecture

Interactive Implementation

visualSearchAI.ts
import { GoogleGenAI } from "@google/genai";

const ai = new GoogleGenAI({ apiKey: process.env.GEMINI_API_KEY! });

export const analyzeImageForVisualSearch = async (buffer: Buffer, mimeType: string) => {
  const response = await ai.models.generateContent({
    model: "gemini-2.5-flash",
    contents: [
      bufferToGenerativePart(buffer, mimeType),
      "Analyze this fashion image and return category, gender, and dominant color hex."
    ],
    config: {
      responseMimeType: "application/json",
      responseSchema: ANALYSIS_SCHEMA,
    },
  });

  return JSON.parse(response.text);
};
04 • Key Features

Core Highlights

AI Visual Search

Upload any fashion image or snapshot to instantly find matching clothing items by category, color, and style.

Automated Gemini Tagging

Zero-shot catalog metadata generation enforcing structured JSON schemas for automated categorization.

Multi-Variant Inventory Engine

Comprehensive product management handling dynamic sizes, colors, SKUs, and real-time stock validation.

Auth & Admin Dashboard

Clerk-authenticated user management with role-based access control and live order tracking.