Build with AI Line
From neural-network intuition to calling model APIs, retrieval over your own documents, simple agents and evaluation.
Intermediate → Advanced14–18 hours7 stationsFree
Who it's for: Developers, technical founders and curious builders who want to ship AI features or small AI products.
What you'll be able to do
- Explain gradient descent and attention at an intuitive level
- Call a model API and handle errors and costs
- Build retrieval-augmented generation over your own files
- Write evaluations that catch regressions
The route
Station 1: Neural networks, visually
Neurons, weights and layers through a visual explanation before any code.
Station 2: How models learn
Loss, gradient descent and backpropagation as one connected story.
Station 3: Transformers and attention
Why attention let models handle long text, and what that means for prompts and context windows.
Station 4: Calling a model API
Requests, parameters, streaming, rate limits and keeping costs predictable.
Station 5: Retrieval over your documents
Chunking, embeddings and search so answers cite your own sources.
Station 6: Tools and simple agents
Let a model call functions, check its own work, and know when to hand back to a person.
Station 7: Evaluating AI features
Test sets, rubrics and regression checks so quality doesn't silently drift.
Read along
- How large language models work: tokens, training and prediction · 10 min
- Retrieval-augmented generation (RAG) explained · 9 min
- How to evaluate AI outputs: test sets, rubrics and regression checks · 8 min
Watch along
But what is a neural network?
3Blue1Brown · Build with AI
A visual first look at neurons, layers and weights. The clearest starting point for intuition.
Gradient descent, how neural networks learn
3Blue1Brown · Build with AI
How a network adjusts itself to reduce error, shown as rolling downhill on a landscape.
What is backpropagation really doing?
3Blue1Brown · Build with AI
The intuition behind how each weight learns its share of the blame.
Transformers, the tech behind LLMs
3Blue1Brown · Build with AI
Tokens, embeddings and the transformer pipeline explained visually.
Attention in transformers, step by step
3Blue1Brown · Build with AI
How attention lets each word use context from the rest of the text.
Let's build GPT: from scratch, in code
Andrej Karpathy · Build with AI
Code a small GPT step by step. For developers who want to see every moving part.
Machine Learning for Everybody – Full Course
freeCodeCamp.org · Build with AI
A beginner-friendly machine-learning course with code.
How to complete this line
Work through one station per sitting. For each: read the linked guide, watch the video where one is listed, then apply the idea to one real task from your own work and note the result. When you have done all stations, try the matching brief in this month's prompt challenge.