This course gives you the spine - reliability, retrieval, agents, evals, memory - built on one real capstone. Mastery comes from going wider and deeper on your own, and the good news is that almost everything you need to master AI engineering is free; you only pay for API usage. This module is a curated map of the best free resources, organized by area, plus how to actually use them.

How to learn this well (read this first)

• Build, do not just watch: every resource should end in a shipped project. The three projects that map directly to what employers want in 2026 are a RAG system over your own docs with retrieval evals, an agent with tool use and a failure-mode eval, and a reusable eval pipeline with a golden set and a validated LLM-as-judge.
• Read other people's code: mature open-source LLM projects teach patterns no tutorial does.
• Invest early in evals and context engineering - the gap between a demo and a production system almost always lives there.

Foundations: how LLMs actually work

• Andrej Karpathy - Neural Networks: Zero to Hero - the single best free resource on what is happening inside these models, taught from first principles.
• 3Blue1Brown - neural network and transformer series - visual intuition on attention and how transformers work.

Prompting and context engineering

• The Prompt Engineering Guide - the most complete free reference on prompting techniques.
• OpenAI Cookbook - hands-on recipes with real code for prompting, structured outputs, and more.
• Anthropic Cookbooks - the same for Claude, strong on tool use and agents.

RAG and retrieval

• Pinecone Learn - the clearest free curriculum on embeddings, chunking, hybrid search, and RAG evaluation.
• Hugging Face Learn - free courses on LLMs and NLP, strong on the open-source side.

Agents and orchestration

• Anthropic - Building Effective Agents - the canonical short read on what an agent is and when to use one.
• LangChain Academy - Intro to LangGraph - free structured course from the maintainers.
• LangChain YouTube - idiomatic agent, RAG, and eval patterns from the core team.
• Model Context Protocol and A2A protocol - the two protocols connecting the agent world.

Evals (the most underrated skill)

• Hamel Husain's blog - definitive free writing on error analysis, LLM-as-judge, and building real eval systems.
• Langfuse docs - tracing and eval tooling in practice.

Depth and systems thinking

• Chip Huyen - the best single framing of AI engineering as systems design.
• Eugene Yan's blog - practical essays on building and evaluating LLM systems in production.
• learn-ai-engineering (GitHub) - curated free-resource collection that stays current.

Staying current (the field rots fast)

• Latent Space - newsletter and podcast tracking the AI-engineering field week to week.
• Simon Willison's blog - sharp, frequent, hands-on notes on what is new and whether it matters.
• Follow the official docs of the tools you use - they change monthly, and the changelog is often the best signal.

How to sequence it

Start with Karpathy for intuition and the Prompt Engineering Guide for the core skill. Build a RAG project using Pinecone Learn. Build an agent using the Anthropic agents piece and LangChain Academy. Then, before you feel ready, build an eval pipeline with Hamel's material - because that is the skill that separates people who can demo from people who can ship. Keep Latent Space and Simon Willison in your weekly rotation. By roughly month three of consistent building, you should be able to ship a small, evaluated AI feature into a real product - which is the bar that gets you hired.