A two-section course

From data to leaderboard: two machine learning projects, explained

Walk through two complete Kaggle-style projects end to end. Each section covers the problem, the data, the modelling choices, and the results. You get figures, short videos, a slide deck, quizzes, and interactive demos along the way.

Section 1 · Deep learning

Industrial surface defect detection

Segment four kinds of defects on steel-surface images with U-Net models, then tune thresholds and post-processing to maximise Dice.

Start section 1 →

Section 2 · Machine learning

Workspace price-tier prediction

Predict one of four price tiers for workspace listings from tabular data with gradient boosting, target encoding, and careful cross-validation for macro F1.

Start section 2 →

How to use this course

Take the sections in either order. They don't depend on each other.
Each section opens with an overview page that has a short video and a link to its slide deck.
Look for ▶ Try it boxes: they're interactive demos, several of them running on the projects' real validation data.
Answer the quiz at the end of each lesson by clicking an option. A lesson is marked as read once you scroll to the bottom.
Switch between light and dark themes with the Theme button.

Prerequisites

You should be comfortable reading Python and know what training and validation sets are. Section 1 assumes you've met convolutional networks; section 2 assumes you've met decision trees.