AI Engineering Guide

What this guide is

This guide is an introductory course on the concepts of AI Engineering and agentic AI: the vocabulary and ideas you need to build applications that use language models (LLMs) reliably. Each module explains a group of concepts, each one in two or three paragraphs at most.

It is a map, not a treatise. The goal is for you to know what each thing is, what it is for and when you don't need it. From there, every concept deserves to be explored further on your own, through documentation, articles and, above all, putting it into practice.

Who it is for

For developers and technical profiles who already know how to code and have consumed an API or two, and who want to make the leap to integrating AI into real applications. You don't need to know maths or to have ever trained a model: the guide focuses on how to use models, not on how they are built.

The concepts are independent of language and framework. When something needs to be made concrete, it refers to things you will find in any SDK or provider, whether OpenAI, Anthropic, Google or an open model.

How to use it

The modules follow a deliberate order: each one builds on the previous one, from LLM fundamentals up to multi-agent systems. Module 11 is the exception: it covers background concepts you can read at any time.

For each module, read the concepts, copy the exam prompt at the end and paste it into your AI (ChatGPT, Claude, Gemini…). It will give you a 20-question test and then suggest a hands-on exercise on your own application. A good criterion for considering a module learned: you can explain it without notes, you have applied it to something real and you know when not to use it.

The running example: a customer support application

To avoid talking in the abstract, every module shares an example application: a customer support application for an online shop, with tickets, customers, orders and a knowledge base (frequently asked questions and return policies).

It is a universal enough case for every concept to have a natural application in it: summarising tickets, answering from the knowledge base, checking the status of an order, defending against a malicious ticket or building an agent that reviews the backlog. If you have an application of your own, the exam prompt will adapt the exercise to it.

Author

I'm Dani Pérez, a software engineer since 2004. This guide grew out of the need for continuous learning, intrinsic to our profession.

The "From practice" boxes in each module collect lessons from applying these concepts in real projects: the things that aren't in the documentation and that you usually learn by getting them wrong.