Picko is a comprehensive recommendation platform that helps users discover movies, TV shows, and books tailored to their tastes.
Overview
The project leverages machine learning techniques including TF-IDF vectorization, cosine similarity, and semantic search to provide personalized recommendations. Available on both iOS and Android app stores, Picko makes content discovery intuitive and engaging.
Key Features
- Smart Recommendations: Uses advanced ML algorithms to suggest content based on user preferences
- Multi-Platform: Deployed on iOS and Android using React Native and Expo
- Large-Scale Database: Over 300,000 items efficiently managed with SQL
- High Performance: Built with FastAPI backend for fast response times
- Cloud Infrastructure: Deployed on AWS with Vercel for optimal performance
Technical Implementation
The recommendation engine combines multiple techniques:
- TF-IDF Vectorization: Analyzes content metadata to understand similarity
- Cosine Similarity: Measures similarity between items
- Semantic Search: Understands user intent beyond keyword matching
The backend uses SQLModel and SQLAlchemy to handle high concurrency and ensure efficient database operations.
Impact
Picko has been successfully deployed to app stores and serves users looking for personalized content recommendations across multiple media types.