Machine Learning A–Z: Ten Real Projects
Churn, pricing, fraud, forecasting — a portfolio in one course.
Taught by Rohan Maharjan · ML Engineer
What you'll learn
- Run a complete machine learning workflow from problem framing to a reported test metric on ten different datasets
- Build regression, classification, time-series, clustering, recommendation, text and image models with the right library for each
- Handle imbalanced data with resampling, class weights and threshold tuning, and evaluate with precision-recall metrics
- Tune gradient-boosted models with cross-validation and explain their predictions using SHAP
- Package a trained model as a FastAPI service with input validation, logging and drift checks
- Present each project as a portfolio piece with a clear README, results table and honest limitations
About this course
The fastest way to become employable in machine learning is to finish real projects end to end, and this course walks you through ten of them. You will predict house prices, flag customers about to churn, catch fraud in heavily imbalanced data, forecast weekly sales, segment customers, build a recommendation engine, classify sentiment in reviews, recognise images with transfer learning, explain a loan-default model with SHAP and finally package a model as a live API with monitoring. Each project repeats the same professional workflow, from framing and data cleaning through feature engineering, modelling, evaluation and hand-off, so by the tenth it is second nature. Tools include pandas, scikit-learn, XGBoost, Prophet, Keras, SHAP and FastAPI. The course is for developers and analysts who know Python and want a portfolio of machine learning projects they can show in interviews. Basic familiarity with pandas and a first machine learning course is assumed.
Course content
8 lessons- 1The Workflow and Project 1: Pricing With RegressionLearn the seven-step workflow you will repeat ten times, then apply it to predicting house prices.Preview
- 2Project 2: Churn Prediction With Gradient BoostingPredict which telecom customers will leave, engineer behavioural features and tune XGBoost for business value.
- 3Project 3: Fraud Detection on Imbalanced DataCatch fraudulent card transactions where positives are 0.2 percent, using class weights, resampling and PR curves.
- 4Project 4: Sales ForecastingForecast weekly store sales with a seasonal baseline, Prophet and a lag-feature gradient boosting model.
- 5Projects 5 and 6: Segmentation and RecommendationsSegment customers with RFM features and k-means, then build a collaborative-filtering recommender on MovieLens.
- 6Projects 7 and 8: Sentiment and Image ClassificationClassify review sentiment with TF-IDF and a fine-tuned transformer, then recognise images with transfer learning in Keras.
- 7Project 9: Explainable Loan Default ScoringBuild a credit-risk model, explain every prediction with SHAP, and check fairness across groups before anyone signs off.
- 8Project 10: Ship It — Packaging, API and MonitoringWrap a trained pipeline in a FastAPI service with validation, logging and drift checks, then assemble the portfolio.