Full-Stack · AI/ML · Real-Time Web App
EZY Notez
Completed · Deployed · Live
Transform your documents, audio & videos into quizzes, flashcards and summaries — powered by AI.
Overview
EZY Notez is a workspace-based learning platform built to solve a real student problem: too much material, too little time to revise. Each workspace represents one subject; students upload resources in any format and unlock AI features grounded entirely in their own content. It combines multi-format ingestion, LLM-powered generation, vector-based retrieval, and real-time multiplayer collaboration into a single production-deployed system across three independently scalable services.
Key features
Resource Pipeline
Ingest PDF, PPTX, audio (Whisper) & YouTube transcripts.
AI Summarization
Bullet, short or detailed (OpenRouter LLM).
Flashcard Generation
Extractive NLP pipeline (NLTK + TF-IDF).
Quiz Generator
MCQ, scenario & mixed types, with server-side scoring.
Chattie
RAG chatbot grounded in resources (Gemini embeddings + pgvector).
Study Rooms
Real-time multiplayer quizzes with WebRTC voice chat.
Tech stack
Frontend
- Next.js
- TypeScript
- Tailwind CSS
- shadcn/ui
- Zustand
Backend
- Express.js (REST API)
ML Service
- FastAPI (Python microservice)
Database
- Supabase
- PostgreSQL
- pgvector
- Auth
- Realtime
- RLS
AI/ML
- OpenRouter (Llama 3.1)
- Google Gemini
- Whisper
- NLTK
Real-Time
- Supabase Realtime
- Socket.IO
- WebRTC
Infra
- Vercel
- Railway
- Docker
- GitHub Actions
Testing
- Playwright (E2E)
- Jest + Supertest
- pytest
Architecture
Three-service architecture separated by runtime boundary (Node vs Python). Long-running AI inference is offloaded to a FastAPI microservice so the Express API never blocks. RAG retrieval uses pgvector cosine similarity over Gemini embeddings.
Highlights
- 5 AI-powered features in one platform
- 3 independently deployable services
- 51 end-to-end tests (Playwright) + unit/integration coverage
- Production CI/CD with GitHub Actions
- Built solo, end to end (design → AI → real-time → deployment)
