AI for developers

Suggestions in the editor, a support chatbot, search by meaning, an agent that edits files: developers call models every day now. From the outside a model is a black box, and you can't debug a black box. This course opens it: first how a model learns at all, then how to build an app with a language model that doesn't break and doesn't lie to the customer.

Choose the language of the code

JavaC#PythonJavaScript

What you learn

The course has two parts.

Part 1: how a model learns

• How a model learns: rules versus examples, a line through the points, the error, a step down the slope, classification, evaluation and the model that memorises.
• A neural network from the inside: a neuron, layers, how the error travels back and the whole training loop, up to a network that recognises digits.

Part 2: apps with language models

• How a language model works: tokens, the next token and temperature, text as a vector, the context window and the model's limits.
• Apps with LLMs: calling an API, the prompt as code, JSON that follows a schema, tools, a test set, cost and speed.
• Semantic search and RAG: document vectors, chunking, answers with a source and checking that it finds the right thing.
• Agents: the plan, act, observe loop, prompt injection and the guards around the model.
• Project: the support assistant: all of it together in one app.

How you learn here

Every lesson starts from a problem in an online shop: how long the order will take, which payment is fraud, whether a review is good or bad, what the support assistant should answer. Then the idea is shown with a scene, synced to the code: the line settling between the points, the ball rolling down the slope, the signal passing through the network.

The text is the same for the four languages, and the code is in the language you choose, with no libraries. Regression and the neural network are loops and arrays written by hand, so you can see every number. The machine learning industry works in Python, and apps with language models are written in every language. Real API calls are examples you run at your place, with your own key.

What you need to know

One of the four languages, at least up to "Collections". The maths is from school: a straight line, a percentage, an average. The derivative is explained as "which way is down and how steep". "Software engineering" is not required, and where it helps (security, architecture), the lesson points to it.