Courses / Data and analytics tools
Machine learning
Build and evaluate models with scikit-learn in a working notebook.
Projects, case studies and mock exams
Build and evaluate a reminder-priority model without changing appointment access.
6 steps · 18 minProject · IntermediateCustomer segmentation projectBuild customer groups and an indexed behavioral summary.
8 steps · 40 minProject · IntermediateMembership churn projectBuild a held-out churn evaluation and score new customer records.
8 steps · 40 minCase study · IntermediateRetail metric choice caseChoose a review model from positive-case coverage and review capacity.
6 steps · 18 minProject · AdvancedFraud review projectSet a fraud review threshold for Nacre Payments when a missed fraud costs eight times as much as a false alarm.
8 steps · 40 minCase study · AdvancedLogistics model diagnosisA delivery model looks accurate on its training data. Work out how it will perform on new deliveries and what to report.
6 steps · 30 minProject · ExpertDemand forecasting projectForecast demand with past-only features, a time-ordered holdout and a baseline to beat.
8 steps · 40 minCase study · ExpertMedia and travel launch reviewReview a media and travel model before launch, starting from the outcome evidence that is available.
6 steps · 30 minMock exam · ExpertModeling job test AA timed modeling test on chronological holdouts, lagged features, grouped validation, drift and an expected-cost threshold.
12 questions · 40 minMock exam · ExpertModeling job test BA second timed modeling test on rolling folds, unseen entities, nested evaluation, rate shifts, rollback and subgroup recall.
12 questions · 45 minWork through the modeling workflow in a real Python notebook. You start with features, targets, training and test splits and baselines. Then you build regression and classification models, reusable preprocessing pipelines and model selection experiments. Later work covers clustering, imbalance, calibrated probabilities, interpretation and chronological evaluation. Original projects ask you to score churn, forecast demand, investigate fraud and describe customer segments. Every task checks your fitted objects and results on a second data fixture. You finish with interview practice and two timed job tests.
Interview practice
Explain what you've learned in a full-screen interview. Speak or type, then review your answers against the rubric for your level.
Interview loops by jobBeginner
The modeling workflow
Frame a decision, identify predictors and define the outcome.
Evaluation splits
Protect validation and test observations from training information.
Regression and baselines
Compare a fitted model with a baseline using held-out errors.
Classification decisions
Fit a binary classifier and measure the consequences of its errors.
Intermediate
Scaling and regularization
Fit transformations on training rows and control coefficient complexity.
Nonlinear estimators
Compare trees, forest averaging and sequential boosting.
Neighbors and customer segments
Use distances for classification and meaningful customer groupings.
Reusable feature preparation
Handle missing values and new categories inside fitted pipelines.
Advanced
Model selection
Build features and compare candidates with validation folds.
Rare events and ranking
Evaluate imbalanced outcomes and choose operational thresholds.
Density and dimensionality
Inspect density clusters and lower-dimensional projections.
Probabilities and interpretation
Calibrate probabilities and inspect predictive contributions without claiming causation.
Expert
Chronological prediction
Validate forecasts with lagged features and strictly later holdouts.
Dependent observations
Keep groups apart and separate model selection from evaluation.
Model operations
Measure drift and mature outcomes before choosing a response.
Model launch decisions
Combine error costs, subgroup evidence and timed assessment work.