🤖Pro

AI & Machine Learning Fundamentals

Master the foundations of AI and machine learning from scratch. Build linear regression, logistic regression, neural networks, CNNs, and NLP pipelines using only NumPy — no frameworks, just understanding.

11 modules 73 lessons ~18h AI voice coach

Start Learning — Pro

7-day free Pro trial included

Course Outline

NumPy Foundations for ML

6 lessons

Master NumPy arrays, vectorized operations, broadcasting, and linear algebra primitives that underpin every ML algorithm in this course.

Why NumPy? Vectors, Matrices, and Speed

Array Creation, Indexing, and Slicing

Vectorized Operations and Broadcasting

Linear Algebra with NumPy

Random Numbers, Seeds, and Distributions

Checkpoint: Build a Mini Data Pipeline

Math Foundations for Machine Learning

6 lessons

Build intuition for the calculus, linear algebra, and probability concepts that appear in every ML derivation — gradients, eigenvectors, Bayes' theorem.

Derivatives and the Chain Rule

Partial Derivatives and Gradient Vectors

Matrix Calculus: Jacobians and Hessians

Probability, Distributions, and Bayes' Theorem

Eigenvalues, Eigenvectors, and SVD Intuition

Checkpoint: Numerical Gradient Checker

Data Preprocessing and Feature Engineering

6 lessons

Transform raw data into ML-ready form: handle missing values, encode categoricals, scale features, detect outliers, and split datasets properly.

Loading and Exploring Datasets

Handling Missing Values and Outliers

Feature Scaling: Normalization vs Standardization

Encoding Categorical Variables

Train, Validation, and Test Splits

Checkpoint: Full Preprocessing Pipeline

Linear Regression from Scratch

7 lessons

Derive and implement ordinary least squares and gradient descent regression using only NumPy. Understand bias-variance tradeoff and regularization.

The Regression Problem: Fitting a Line to Data

Mean Squared Error: The Loss Function

The Normal Equation: Closed-Form Solution

Gradient Descent for Linear Regression

Ridge and Lasso Regularization

Polynomial Features and the Bias-Variance Tradeoff

Checkpoint: Predict Housing Prices

Logistic Regression and Binary Classification

7 lessons

Build a binary classifier from scratch using the sigmoid function, cross-entropy loss, and gradient descent. Evaluate with precision, recall, and ROC curves.

The Classification Problem: From Regression to Probabilities

Sigmoid Activation and Cross-Entropy Loss

Training Logistic Regression with Gradient Descent

Decision Boundaries and Threshold Tuning

Evaluation Metrics: Accuracy, Precision, Recall, F1, ROC-AUC

Multiclass Classification with Softmax

Checkpoint: Spam Classifier

Neural Networks from Scratch

7 lessons

Build a fully connected feedforward neural network using only NumPy. Implement forward propagation, backpropagation, and train on real data.

From Perceptron to Multi-Layer Network

Activation Functions: ReLU, Sigmoid, Tanh, and Softmax

Forward Propagation: Computing Predictions

Backpropagation: Computing Gradients Layer by Layer

Weight Initialization: Xavier and He

The Training Loop: Mini-Batch Gradient Descent

Checkpoint: XOR and MNIST Digit Classifier

Training Deep Networks: Optimization and Regularization

7 lessons

Move beyond vanilla SGD with Adam, momentum, and learning rate schedules. Prevent overfitting with dropout, batch norm, and early stopping.

SGD with Momentum and Nesterov Acceleration

Adaptive Optimizers: RMSProp and Adam

Learning Rate Schedules and Warmup

Dropout: Regularization by Random Deactivation

Batch Normalization: Normalizing Hidden Layers

Early Stopping and Validation Curves

Checkpoint: Ablation Study on MNIST

Convolutional Neural Networks

7 lessons

Master the convolution operation, pooling, and CNN architectures. Build and train a CNN from scratch, then explore transfer learning for practical image classification.

The Convolution Operation from Scratch

Pooling, Stride, and Receptive Fields

Classic CNN Architectures: LeNet, VGG, and ResNet

Transfer Learning: Fine-Tuning Pretrained Models

Data Augmentation for Image Models

From Classification to Detection: YOLO Intuition

Checkpoint: Build a CNN Image Classifier

NLP Fundamentals and Text Pipelines

7 lessons

Process raw text into ML-ready features. Learn tokenization, TF-IDF, word embeddings, and the sequence models (RNN, LSTM) that paved the way for Transformers.

Text Preprocessing and Tokenization

Bag of Words and TF-IDF

Word Embeddings: Word2Vec and GloVe

Recurrent Networks: RNN, LSTM, and GRU

The Attention Mechanism

Transformer Architecture Intuition

Checkpoint: Sentiment Classifier Pipeline

Unsupervised Learning: Clustering and Dimensionality Reduction

6 lessons

Find structure in unlabeled data. Implement K-Means and DBSCAN clustering, reduce dimensions with PCA and t-SNE, and build autoencoders for learned compression.

K-Means Clustering

DBSCAN: Density-Based Clustering

Principal Component Analysis (PCA)

t-SNE: Nonlinear Dimensionality Reduction

Autoencoders: Learned Compression

Checkpoint: Customer Segmentation Pipeline

Model Evaluation, Selection, and Deployment Readiness

7 lessons

Move from a trained model to a production-ready artifact: cross-validation, hyperparameter search, calibration, serialization, and inference pipelines.

K-Fold and Stratified Cross-Validation

Grid Search and Random Search

Probability Calibration and Reliability Diagrams

Serializing and Loading Models with NumPy

Building a Reproducible Inference Pipeline

ML System Design Primer: From Notebook to Production

Capstone: End-to-End ML Project