
Phase 0: Tools
Python from your first line, plus the terminal and git, ending in an ad-copy reviewer, with all 44 lessons free.
What happens after pretraining, from reward models to quantization, ending with an assistant you tune and test.
This phase covers what happens to a model after pretraining. You train a reward model and store weights in fewer bits. It ends with an assistant you tune and test. 22 lessons, about 16 hours. The first lesson is free: Base models vs assistants. Start free at indexaiacademy.com.
All 38 videos, by what they're for.

Python from your first line, plus the terminal and git, ending in an ad-copy reviewer, with all 44 lessons free.

Building with language models: a help-center bot that answers from documents, then a research agent that cites sources.

The math that models run on, from gradient descent to cross-entropy, the loss language models are trained on.

Learning from data and knowing when it worked, from fitting a line to predicting which customers will leave.

Neural networks and how to train them: backpropagation you write yourself, PyTorch, and a classifier tested fairly.

How transformers work: attention written line by line, then a tiny GPT trained in your browser.

How to build software well, from data structures to tests, with a capstone command-line to-do app backed by a database.

The servers your code talks to, from raw HTTP to a server with no framework and a complete API for product reviews.

MCP servers, clients and hosts, from JSON-RPC 2.0 to a tested server for a shop's support team.

The code around a model that makes it an agent, ending with a coding agent that stops when the tests pass.

Shipping and running real systems, code review and A/B tests, ending with an AI service for support tickets.
Never trained an AI? Teach a small one to write, right in your browser. It takes about 20 seconds and needs no account. Teach it to write