Machine Learning Crash Course

Who this is for: Anyone who knows basic Python (loops, functions, lists) and wants to understand machine learning by doing it - not by reading about it.

Approach: You train your first model in the second cell of Module 1, before we explain anything. That's intentional. Results first, theory second.

After this course you will:

• Train and evaluate machine learning models on real data
• Understand what happens inside model.fit() - loss, gradients, epochs
• Know which model to reach for and why
• Handle messy real-world data
• Build a complete end-to-end ML project
• (With extension modules) tune hyperparameters, engineer features, deploy to production

Modules

Core (10-14 hours)

#	Folder	Notebook	What you build
1	01_First_Model	predicting_house_prices.ipynb	A model that predicts California house prices
2	02_How_It_Learns	how_a_model_learns.ipynb	Linear & logistic regression from scratch - see loss curves, gradient descent
3	03_Does_It_Work	evaluating_your_model.ipynb	Tools to know if your model is actually good
4	04_Better_Models	classification_with_real_data.ipynb	5 classifiers on the Titanic dataset
5	05_Messy_Data	cleaning_and_preparing_data.ipynb	A complete data preparation pipeline
6	06_Your_Project	end_to_end_project.ipynb (+ unsupervised_intro.ipynb)	A heart disease classifier, start to finish

Extension (optional, 6-10 hours)

#	Folder	Notebook	What you build
7	07_Model_Selection	01_cross_validation_and_tuning.ipynb	Systematic hyperparameter tuning
8	08_Practical_ML	01_feature_engineering_and_pipelines.ipynb	Production-ready feature pipelines
9	09_Capstones	(README only)	Project ideas to apply everything
10	10_Deployment_For_ML	01_deployment_basics.ipynb	Ship classical ML to production

Each chapter folder has a README.md - a deep concept reference (200-500 lines) covering everything the notebook touches.

Estimated time

• Core (1-6): 10-14 hours total.
• Extension (7-10): 6-10 hours additional.

Prerequisites

• Python basics: variables, loops, functions, lists.
• No math beyond high-school level.
• No prior ML knowledge.

Tech stack

All notebooks run in Google Colab - no local setup needed.

scikit-learn  models, preprocessing, evaluation
pandas        data manipulation
matplotlib    visualisation
seaborn       statistical charts
xgboost       industry-standard gradient boosting
joblib        saving trained models

How each notebook works

Every notebook follows the same rhythm:

1. Run something impressive - before any explanation
2. "What just happened?" - explain what you just saw
3. Just-in-time theory - markdown cells before each code cell explain what's about to happen
4. Failure modes - see what goes wrong and why
5. Exercises - try it yourself

Companion README.md in each chapter goes deep on every concept the notebook touches.