Beginner
From what AI is to a working assistant that answers from your own documents. Read the levels in order, because each one assumes the ones before it. The track ends with a guided project.
Paid track · coming soonAI Basics
No prerequisites. Written for someone who has only used a chat assistant.
After this level you can decide whether a problem calls for generative AI at all, or for ordinary software, search or classic machine learning.
How AI Works
The machinery, one idea per page.
After this level you can decide how much text you can send a model at once, and what that costs you in tokens.
Talking to a Model
Getting a usable answer out.
After this level you can decide whether a better prompt, a lower temperature, step-by-step reasoning or a fixed output format will fix the answers you're getting.
Giving a Model Your Own Data
How a model answers from your documents, built up one part at a time.
After this level you can decide whether your documents need retrieval, and how to split, index and search them: by keyword, by meaning, or both.
- Why a model doesn't know your data
- Semantic ChunkingFree
- Vector IndexFree
- BM25Free
- Hybrid SearchFree
- Putting RAG together
Making a Model Do Things
How a model goes from answering to acting.
After this level you can decide whether a task should be one model call or an agent with tools, and which tools to give it.
Knowing Whether It Works
Measuring quality before users find the problems.
After this level you can decide whether your system works: what to measure, against which test set, and whether a model can do the grading.
Using AI Responsibly
The safety basics every feature needs.
After this level you can decide which safety checks a feature needs before launch, and what people must never paste into it.
Cost and Choices
What it costs, and which knobs to turn.
After this level you can decide which model to use, whether caching or fine-tuning pays for itself, and what a feature will cost per request.