Build AI agents and ML models an analyst can trust.

An interactive textbook in ten chapters, set at a fictional research firm. Run Python in the page, poke the visuals, then take it to a real model in Colab.

Open the book →Jump to Setup

No install. Everything runs in your browser.

10chapters

8Friday sessions

40+interactive visuals

6sessions done

The course, session by session

Six of eight sessions are done. Session 8 is next, Fri Nov 13; chapter 7 is self-paced.

All0 · FoundationsI · AgentsII · MLIII · SystemsIV · Finance

What the series automates

The firm's goal: automate its research workflow with an agent plus ML, with a person approving every result.

0 ✓
Foundations

1 ✓
Agent loop

2 ✓
Context

3 ✓
Memory, RAG

4 ✓
ML I

5 ✓
ML II

6 ✓
Self-learning

7
Choose

8 ← next
Systems

9
Markets

Try one right here

A real widget from chapter 0, running in this page.

/// Interactive · predict the next word

Start with:

What the model thinks comes next

Where this number comes from

Temperature: 1.0 0 always takes the favourite. Higher gives the long shots a better chance.

A toy model: it has read thirteen sentences from the firm's memos and counts which word follows which. A real LLM does the same job over tokens, with a neural network trained on trillions of them. Point at a bar to see how its number is counted; click it to choose that word yourself.

Ready? It takes ten minutes to set up.

Read the firm page, then get a working Colab with an API key.

Open the book →