13.1 Capstone 1 — tabular ML, deployed
A live service that takes an input and returns a calibrated prediction, with a README anyone can follow.
The first capstone assembles everything tabular into a deployed, calibrated prediction service with an honest write-up — the strongest single artifact a portfolio can hold. It opens the capstone module because tabular is where the course's classical spine pays off. The dataset choice is the make-or-break: pick data attached to a real decision, because interviewers probe the decision, and a scrubbed classic dataset has no decision left in it.
Work through these
Pick a real dataset with a real decision attached
The dataset must have a real decision attached, or the project becomes an exercise in scoring rather than in usefulness. Choosing well is most of the value of this capstone.
EDA, feature engineering, model selection, honest CV
The full modelling sequence done honestly, with validation that respects how the data was generated. This is where every technique from the classical machine learning module is exercised at once.
FastAPI service, Docker, deployed publicly
A running service in a container, deployed somewhere a stranger can reach it. A model that lives only in a notebook cannot be shown to anyone.
Write-up: problem, approach, results, limitations
A written account of the problem, the approach, the result and what the model cannot do. The limitations section is what a knowledgeable reader looks at first.
Sign in to keep your progress.
Free resources
We haven't checked most of these for screen reader use yet.
Links last checked 29 Aug 2026.
Stuck here?
Ask a mentor. A real person answers, and they can see exactly which topic you're on. Usually within a couple of working days.
Checking your session…
Topics shown in module order.