Tutorial Series // AI & ML

The Complete AI & ML Tutorial

Machine learning from first principles — regression to neural networks, evaluation, and MLOps. Every topic ships working code.

12 topics ~113 min total read 100% free
  1. 01 What Is Machine Learning, Really? Supervised vs unsupervised learning, the standard ML workflow, and setting up a Python environment for the whole series. 8 min →
  2. 02 NumPy & pandas for Machine Learning The 20% of NumPy and pandas you need for ML — arrays, vectorisation, DataFrames, and turning raw data into feature matrices. 9 min →
  3. 03 Linear Regression from First Principles Your first model — fitting a line with least squares, understanding loss, gradient descent, and reading coefficients. 10 min →
  4. 04 Classification with Logistic Regression Predicting categories instead of numbers — sigmoid, decision thresholds, class probabilities, and a spam classifier. 9 min →
  5. 05 Model Evaluation: Beyond Accuracy Why accuracy lies on imbalanced data — confusion matrices, precision, recall, F1, ROC-AUC, and cross-validation. 10 min →
  6. 06 Overfitting, Regularization & Bias-Variance Why models memorise instead of learn — detecting overfitting, L1/L2 regularization, and the bias-variance trade-off. 9 min →
  7. 07 Decision Trees & Random Forests Trees that split data with if/else questions, why single trees overfit, and how random forests fix them with bagging. 9 min →
  8. 08 Gradient Boosting: XGBoost & Friends The algorithm that wins on tabular data — boosting intuition, XGBoost in practice, early stopping, and tuning that matters. 9 min →
  9. 09 Unsupervised Learning: Clustering & PCA Finding structure without labels — k-means clustering, choosing k, DBSCAN, and dimensionality reduction with PCA. 9 min →
  10. 10 Neural Networks from Scratch Neurons, layers, activations, and backpropagation — building a working two-layer network in raw NumPy. 11 min →
  11. 11 Deep Learning with PyTorch Tensors, autograd, nn.Module, DataLoaders, and a complete training loop — the real-world deep learning workflow. 10 min →
  12. 12 Shipping Models: MLOps Fundamentals From notebook to production — pipelines, model serialization, serving with FastAPI, monitoring, and retraining. 10 min →