Production-Agentic-RAG-Course is an educational project that teaches how to build production-grade retrieval-augmented generation systems through hands-on implementation of a research assistant.
The course addresses the gap between theoretical AI knowledge and practical production systems by teaching RAG development through a structured, week-by-week progression. Rather than jumping directly to vector search, the approach emphasizes building solid keyword search foundations first using BM25, then enhancing with semantic understanding through hybrid retrieval. The system fetches academic papers from arXiv, parses their content, and answers research questions using advanced RAG techniques. The architecture evolves from basic infrastructure setup through data pipelines, keyword search, intelligent chunking, complete RAG pipelines with local LLMs, production monitoring with tracing and caching, and finally agentic RAG with decision-making capabilities and Telegram bot integration.
This course suits developers who want to understand how successful companies build RAG systems in production rather than following AI-first approaches that neglect search fundamentals. It works best for those learning through implementation, as each week builds a functional component of a complete research assistant system. The progression from infrastructure setup through monitoring and agentic features means learners gain both breadth across the full stack and depth in each component. The project uses FastAPI, PostgreSQL, OpenSearch, Docker, Airflow, and LangGraph, providing exposure to industry-standard tools.
The project shows active development with substantial feature additions across multiple weeks, including infrastructure setup, data pipeline automation, search implementations, RAG pipeline completion, monitoring integration, and agentic capabilities. The course includes detailed code walkthroughs and accompanying blog content explaining the architectural decisions and implementation details. The system architecture documentation demonstrates iterative refinement, with visual diagrams showing how components evolve and integrate across the course progression.
