Production Agentic RAG Course: Build an arXiv Paper Curator

Key takeaways#

• Production Agentic RAG Course teaches a full research-assistant build, not a toy vector-search demo.
• The project starts with infrastructure, ingestion, and keyword search before adding chunking, hybrid retrieval, local LLM responses, tracing, Redis caching, LangGraph agents, and a Telegram interface.
• The README requires Docker Desktop, Python 3.12 or newer, uv, at least 8 GB of RAM, and roughly 20 GB of free disk space.
• It is best for builders who want to understand production RAG architecture through weekly releases and code walkthroughs.

What you build#

This course walks through an arXiv Paper Curator: a research assistant that fetches academic papers, parses them, indexes them, and answers research questions with retrieval-augmented generation. The repository explicitly frames the learning path around production foundations. Week 1 sets up Docker, FastAPI, PostgreSQL, OpenSearch, and Airflow. Week 2 adds a data pipeline for arXiv papers. Week 3 focuses on BM25 keyword search, filtering, and relevance scoring. Week 4 adds chunking plus hybrid search. Week 5 completes the RAG loop with a local LLM, streaming responses, and a Gradio interface. Week 6 introduces Langfuse tracing and Redis caching. Week 7 adds LangGraph-based agentic RAG and a Telegram bot.

Why it matters#

Many RAG tutorials skip directly to embeddings and ignore search fundamentals. This course is useful because it keeps classic information retrieval in the critical path, then layers vectors and agents on top. That mirrors how many production teams build reliable systems: deterministic ingestion and search first, AI reasoning second, observability throughout.

Best fit#

Choose this resource if you already know basic Python and want a deeper system-building path. The stack is heavier than a notebook tutorial, but that is the point: you learn service boundaries, queues, indexes, monitoring, and guardrails while still building an AI-native product.