Skip to main content

Welcome to Few-Shot Academy

A free, open-source, chapter-wise curriculum for Generative AI, LLMs, Vector Databases, RAG, and Agents — from "what is a token?" to a working, evaluated agentic RAG system.

Why this exists​

Generative AI is exploding, but most learning paths force a tradeoff: either shallow, video-only "for everyone" content with no hands-on practice, or hands-on labs that assume you already know how to code and are willing to pay for cloud credits. Few-Shot Academy is neither. It's:

  • Truly beginner-friendly — Foundations assumes zero prior AI knowledge.
  • Hands-on at every step — every lesson links to runnable code you execute on your own machine.
  • Local-first and free — vector storage runs locally via ChromaDB, and every Foundations/Intermediate lab works with a free local Ollama model. No credit card required to finish the first two tracks. Prefer a hosted model? Bring your own OpenAI or Anthropic API key instead.
  • One coherent arc — three tracks, each building on the last, ending with you having built and evaluated a real agentic RAG system.

How the curriculum is organized​

  • Foundations — start here if you're new to AI.
  • Intermediate — chunking, retrieval quality, tool use, your first agent.
  • Advanced — multi-agent systems, advanced RAG, guardrails, observability, shipping to production.

Every chapter follows the same shape: Concept → Diagram → Hands-on Lab → Checkpoint → What's Next. The hands-on labs live in the labs/ folder of the companion repository — clone it once, and you're set up for every chapter.

What you'll walk away with​

Each track ends with something you actually built, not just read about:

  • Foundations → a working Q&A bot that answers questions over your own documents, using retrieval-augmented generation you understand piece by piece.
  • Intermediate → a tool-calling assistant, then a multi-tool agent that combines web search, a calculator, and RAG over your own documents, plus the ability to judge whether a RAG or agent system is actually working, not just build one.
  • Advanced → a production-shaped agentic RAG system with evaluation, guardrails, and observability built in, packaged the way a real system would ship.

Whichever track you land on, you'll finish it having built something real, not just read about it.