Skip to content
Back to ProjectsFull-Stack + Cloud

SneakUp

With AI Features (Visual Search, NLP Smart Search, Recommendation Engine)

Production-grade streetwear eCommerce with AI visual search, NLP smart search & real-time order tracking

WEBDESIGNCLOUDAILIVEPRODUCTION
PRODUCTION GRADE
Active
Aug 2025 - Feb 2026

// VIDEO SHOWCASE

01/06AI Visual Search Demo
COMING SOON
COMING SOON
COMING SOON
COMING SOON
COMING SOON
COMING SOON
MISSION STORY:

FROM PORTFOLIO PROJECT TO PRODUCTION GRADE

SneakUp is a production-grade, full-stack streetwear and sneaker eCommerce platform — my first project engineered to senior-level standards. Originally built as a MERN stack portfolio project, it underwent a systematic 6-phase transformation addressing every dimension of production software: server-side price calculation (preventing payment fraud), JWT refresh token rotation with SHA-256 hashing, 4-tier rate limiting, Zod validation on every input, full-stack testing (Jest, Vitest, Playwright, k6), CI/CD with 5 parallel GitHub Actions jobs, Sentry error tracking, Pino structured logging, and comprehensive documentation (20+ files including architecture decisions, incident runbooks, and security audits). The platform features AI-powered visual search (CLIP + FAISS), NLP smart search, a personalized recommendation engine blending collaborative filtering + content-based similarity + trending, real-time order tracking via Socket.io, Google OAuth, PWA support, and an admin analytics dashboard — all deployed across Vercel, Railway, Hugging Face Spaces, and MongoDB Atlas for ~$0–5/month.

25+API Endpoints
22Frontend Pages
6Database Models
20+Reusable Components
6Email Templates
7.64msp95 Response Time
~48 req/sLoad Test Throughput
0.00%Error Rate
All MitigatedOWASP Top 10
5CI Pipeline Jobs
20+Documentation Files
13ADRs Written

Multi-service architecture with 4 independently deployable services. Backend follows a layered pattern: Controller → Service → Repository with an event-driven side-effect system via typed EventBus. 13 middleware in precise stack order. 6 MongoDB collections with TTL indexes for automatic cleanup.

Frontend SPA

React 19 + Vite 7TAP TO FLIP

22 lazy-loaded pages, TanStack Query for server state, React Hook Form + Zod for validated forms, Socket.io client for real-time updates. Deployed on Vercel Edge CDN.

Backend API

Express 5 + TypeScriptTAP TO FLIP

25+ endpoints across 8 route groups with layered architecture (8 controllers, 7 services, 7 repositories). Stripe payments, JWT auth, 4-tier rate limiting. Deployed on Railway.

AI Service

FastAPI + CLIP + FAISSTAP TO FLIP

Visual search, NLP search, and recommendation engine. CLIP ViT-B/32 encodes images/text to 512-dim vectors, FAISS performs sub-millisecond similarity search. Deployed on Hugging Face Spaces (16GB RAM).

Data Layer

MongoDB Atlas + Upstash RedisTAP TO FLIP

6 collections (Products, Orders, Users, Reviews, UserEvents w/ 90-day TTL, RefreshTokens w/ auto-expire). Redis cache-aside pattern with graceful degradation.

FRONTEND

React 19Vite 7TypeScriptTailwind CSS 4TanStack React QuerySK-0028SK-0095React Router 6SK-0027SK-0025SK-0073SK-0083SK-0069

BACKEND

Express 5TypeScriptSK-0026SK-0095SK-0066SK-0089SK-0055SK-0027SK-0064SK-0093SK-0094SwaggerSK-0080SK-0102

DATABASE

MongoDB AtlasUpstash Redis

CLOUD & DEVOPS

DockerGitHub ActionsSK-0046SK-0044SK-0044Terraform

TOOLS & TESTING

FastAPISK-0101SK-0104PyTorchPythonJestSK-0078SK-0068SK-0079SK-0084SK-0069
PhilippinesREMOTE GLOBALwww.kbuendia.com