Data Science AI ML Course

Lesson 1 • Mathematics and Statistics Primer

Reviews linear algebra, calculus intuition, and probability needed for ML algorithms. Grounds abstract maths in data science applications.

Lesson 2 • Data Types and Structures

Distinguishes structured, semi-structured, and unstructured data formats. Prepares learners to handle diverse real-world datasets.

Lesson 3 • The Data Science Landscape

Defines data science, AI, and ML as distinct but overlapping disciplines. Establishes vocabulary used throughout the course.

Lesson 4 • Setting Up Your Environment

Installs and configures Python, Jupyter, and essential libraries. Ensures every learner has a reproducible workspace.

Lesson 1 • Data Cleaning Techniques

Identifies and resolves missing values, duplicates, and inconsistent formats. Directly improves downstream model accuracy.

Lesson 2 • Feature Engineering Fundamentals

Transforms raw variables into informative features for ML models. Introduces encoding, scaling, and derived feature creation.

Lesson 3 • Exploratory Data Analysis

Uses statistical summaries and visualisations to uncover patterns and anomalies. Informs hypothesis generation before modelling.

Lesson 4 • Data Ingestion and Loading

Loads data from files, APIs, and databases using pandas and SQL connectors. Establishes the entry point for every data pipeline.

Lesson 1 • Model Evaluation and Validation

Applies train-test splits, cross-validation, and performance metrics to assess models. Prevents data leakage and ensures reliable evaluation.

Lesson 2 • Classification Algorithms

Covers logistic regression, decision trees, and k-nearest neighbours for classification tasks. Highlights decision boundaries and probability outputs.

Lesson 3 • Regression Algorithms

Implements linear, polynomial, and regularised regression models. Connects mathematical derivations to scikit-learn implementations.

Lesson 4 • Hyperparameter Tuning

Optimises model performance through grid search and random search strategies. Introduces the concept of a validation set distinct from the test set.

Lesson 1 • Model Interpretability

Uses SHAP values and permutation importance to explain complex model predictions. Connects interpretability to stakeholder trust and debugging.

Lesson 2 • Bagging and Random Forests

Explains bootstrap aggregation and its variance-reduction effect. Implements random forests and interprets feature importance scores.

Lesson 3 • Support Vector Machines

Derives the maximum-margin classifier and kernel trick for nonlinear boundaries. Applies SVMs to classification and regression problems.

Lesson 4 • Stacking and Blending

Combines diverse base models through meta-learners to reduce generalisation error. Introduces proper stacking pipelines to prevent leakage.

Lesson 1 • Anomaly and Outlier Detection

Detects rare events using isolation forests, autoencoders, and statistical methods. Applies techniques to fraud detection and quality control scenarios.

Lesson 2 • Dimensionality Reduction

Reduces feature space with PCA, t-SNE, and UMAP while preserving structure. Enables visualisation and speeds up downstream modelling.

Lesson 3 • Clustering Algorithms

Implements k-means, hierarchical, and DBSCAN clustering methods. Evaluates cluster quality with silhouette scores and elbow analysis.

Lesson 1 • Backpropagation and Optimisation

Derives backpropagation and connects it to gradient descent variants. Covers optimisers like Adam, RMSProp, and learning rate scheduling.

Lesson 2 • Regularisation and Training Best Practices

Applies dropout, batch normalisation, and early stopping to prevent overfitting. Establishes reliable training workflows for production-grade models.

Lesson 3 • Recurrent Neural Networks and LSTMs

Models sequential data with RNNs, LSTMs, and GRUs. Applies architectures to time-series forecasting and text sequence tasks.

Lesson 4 • Convolutional Neural Networks

Builds CNNs for image classification using convolution, pooling, and fully connected layers. Introduces transfer learning with pretrained models.

Lesson 1 • Word Embeddings

Trains and uses dense vector representations with Word2Vec, GloVe, and FastText. Demonstrates how semantic similarity is encoded in embedding space.

Lesson 2 • Classical NLP Tasks

Implements sentiment analysis, named entity recognition, and text classification. Connects traditional ML models to NLP feature pipelines.

Lesson 3 • Fine-Tuning Pretrained Language Models

Fine-tunes BERT and GPT-style models on downstream NLP tasks using Hugging Face. Covers prompt engineering and evaluation of language model outputs.

Lesson 4 • Text Preprocessing and Representation

Cleans and tokenises raw text, then converts it to numerical representations. Establishes the input pipeline for all NLP models.

Lesson 1 • Model Monitoring and Drift Detection

Tracks prediction quality, data drift, and concept drift in live systems. Triggers retraining pipelines when performance degrades.

Lesson 2 • Model Packaging and Serving

Packages models as REST APIs using Flask and FastAPI, then containerises with Docker. Enables scalable, language-agnostic model consumption.

Lesson 3 • ML Pipelines and Workflow Automation

Designs reproducible end-to-end pipelines using scikit-learn Pipeline and orchestration tools. Eliminates manual steps that cause production failures.

Lesson 4 • MLOps Principles and CI/CD

Applies DevOps practices to ML: versioning data, code, and models together. Implements CI/CD pipelines that test and deploy models automatically.