The Mother of AI Project
A hands-on course that builds a production-grade retrieval-augmented-generation research assistant from scratch, called the arXiv Paper Curator, teaching keyword search first and then hybrid and agentic retrieval.
The Mother of AI Project is a learner-focused course whose first phase, the arXiv Paper Curator, builds a complete research-assistant system that fetches academic papers, understands their content, and answers research questions using retrieval-augmented generation (RAG). It teaches RAG the way the course describes production systems being built: keyword search first, then vector search added on top for hybrid retrieval, rather than starting from vectors alone.
Over seven weeks, it builds infrastructure with Docker, FastAPI, PostgreSQL, OpenSearch and Airflow; an automated pipeline that fetches and parses papers from arXiv; production BM25 keyword search with filtering and relevance scoring; intelligent chunking and hybrid search combining keywords with semantic similarity; a full RAG pipeline with a local LLM, streaming responses and a Gradio interface; production monitoring with Langfuse tracing and Redis caching; and an agentic RAG system built with LangGraph, including a Telegram bot for mobile access, with agents that grade document relevance and rewrite queries adaptively.
It is written in Python under the MIT license.