Hackathon Showcase
Video-RAG Learning Assistant
Visual-Audio Learning from YouTube's Best Content
YouTube Video
Project Description
Visual-Audio Learning from YouTube’s Best Content levveraging video-rag techniques to index mass amount of videos and serve them in tiktok stype clips for self-guided learning.
Key pieces of work on the backend
- Multimodal - Video + Audio + Text in same embedding space
- Knowledge Graph - Structured concept relationships
- Hybrid Retrieval - Three retrieval paths + RRF fusion
- Educational Focus - Socratic tutoring, progress tracking
- Apple Silicon - MLX optimized for local inference
Team
Products & Tools
AI Tinkerers
OpenAI
Redis for AI