AI/MLLLM IntegrationFull-Stack Application
Open Recommender — Open Source AI Recommendation System for Videos & Articles
Self-hosted AI that learns your interests for timeless content
Open source AI-powered recommendation system for YouTube videos and articles (Dec 2023 – Feb 2024). Uses LLMs to learn from your interests and surface timeless, learning-focused content. Self-hosted solution built with TypeScript and Python, featuring tRPC for type-safe APIs and Prisma for database management.
Tech Stack
Node.jsTypeScriptReacttRPCPrismaPythonLangChainOpenAIGit/GitHub
Technical Highlights
- Backend built with Node.js, tRPC for end-to-end type safety, and Prisma ORM for database operations
- Python workers for AI processing with Modelfusion, LangChain, and OpenAI-compatible APIs
- React + TypeScript frontend with seamless integration to type-safe backend APIs
- Content ingestion from YouTube (via yt-dlp), Twitter/X (optional), and web search
Challenges & Solutions
- Learning user preferences → Implemented LLM-based interest modeling that adapts to user behavior over time
- Content quality filtering → Built AI pipeline to prioritize timeless, educational content over viral trends
Gallery