Every path in the program, from Python foundations to production AI systems. Each module builds on the last toward shipping a real, end-to-end ML product.
A welcome letter, your environment, and a live portfolio site. Free for everyone — no account required.
Eleven modules that take you from maintainable Python to a deployed, monitored ML product you build yourself.
Move from writing Python scripts to building maintainable ML software: importable packages, typed data contracts, and reproducible environments.
Module 2 · Coming soonTurn messy data into reliable features: one deterministic transform that produces identical output at train time and at serve time.
Module 3 · Coming soonUnderstand when a model result is trustworthy: imbalance, leakage, the metrics that lie, and splits that estimate the future.
Module 4 · Coming soonTrain models that solve real problems — and know when a better leaderboard number is a worse product: trees, XGBoost, calibration, serialization.
Module 5 · Coming soonMake the ML system run anywhere: the same model, the same dependencies, the same result, on a machine that has never seen the code.
Module 6 · Coming soonTurn a model into a service other code can call: schema-validated, batch-capable, honest about its latency, and a stable contract for its consumers.
Module 7 · Coming soonPut the model in a non-ML human's hands: a minimal dashboard that calls the API and shows a prediction someone can act on — without lying about confidence.
Module 8 · Coming soonMake the system publicly usable: a stranger can reach it at a real URL, over TLS, with one-step rollback and a bounded, stated cost.
Module 9 · Coming soonMake the infrastructure reproducible the way Docker made the app reproducible: declarative, versioned, reviewable infra — the principle, not Terraform syntax.
Module 10 · Coming soonKnow when your system is failing — including the case where it returns 200 OK and is quietly wrong: metrics, drift, alerting, and silent failures.
Module 11 · Coming soonBuild systems that improve safely: CI that tests data and models, and a retraining loop that rolls out a new model version without betting the whole system on it.