AI and Machine Learning

A rigorous introduction to the linear algebra, calculus, and probability theory that underpin all machine learning algorithms. Topics include vector operations, matrix multiplication, derivatives, and Bayes' theorem — presented through practical examples relevant to data and AI work.

Comprehensive Python instruction from first principles through to object-oriented design. Students learn data types, control flow, functions, file handling, and the core libraries — NumPy, Pandas, and Matplotlib — used by professional data scientists daily.

Probability distributions, hypothesis testing, correlation, regression, and statistical inference — all taught in the context of real data problems. Students learn to distinguish meaningful patterns from noise and to communicate statistical findings clearly.

Raw data is almost never analysis-ready. This module teaches the full data preparation pipeline: sourcing data from APIs and databases, identifying missing values and outliers, encoding categorical variables, and normalising features for model input.

Students learn to interrogate data visually before building any model. Topics include distribution analysis, correlation matrices, time-series visualisation, and the design of clear, professional charts using Matplotlib and Seaborn.

A conceptual and historical overview of AI — covering symbolic AI, expert systems, the rise of machine learning, and the current era of deep learning and generative models. Students map the AI landscape and identify where different approaches are most and least effective.

The computational building blocks every AI engineer must understand — arrays, trees, hash tables, sorting algorithms, and complexity analysis. Applied to AI contexts including search problems, graph traversal, and efficient data processing.

Introduction to relational databases and structured query language. Students query, filter, join, and aggregate data from relational databases — an essential foundation for working with training data at scale in professional AI environments.

In-depth study of linear regression, logistic regression, support vector machines, k-nearest neighbours, decision trees, and ensemble methods including random forests and gradient boosting. Every algorithm is implemented from scratch before being applied using Scikit-learn.

K-means clustering, DBSCAN, hierarchical clustering, and association rule mining. Students apply unsupervised methods to customer segmentation, anomaly detection, and recommendation system foundations.

The mathematics of the perceptron, multi-layer feedforward networks, backpropagation, activation functions, and loss functions — taught rigorously before students move to framework-based implementation using TensorFlow and Keras.

The techniques that separate high-performing models from average ones: feature selection, polynomial features, interaction terms, hyperparameter tuning using grid search and random search, and cross-validation strategies that prevent data leakage.

Accuracy, precision, recall, F1 score, ROC-AUC, mean squared error, and confusion matrices — with a deep understanding of when each metric is and is not appropriate. Students learn to evaluate models honestly and communicate results to non-technical stakeholders.

Convolutional neural networks for image classification, recurrent structures for sequential data, and the Keras functional API for building complex model architectures. Students train and evaluate CNNs on standard benchmark datasets.

Professional machine learning practice requires reproducible workflows. This module covers data pipeline design using Scikit-learn Pipelines, model versioning with MLflow, experiment tracking, and the basics of collaborative ML development using Git.

A rigorous examination of how bias enters machine learning systems — through biased training data, biased feature selection, and biased evaluation — and the technical and organisational methods used to identify, measure, and mitigate it, with case studies from healthcare, criminal justice, and finance.

Transfer learning using VGG, ResNet, and EfficientNet. Object detection with YOLO and SSD. Image segmentation with U-Net. Students fine-tune pre-trained models on custom datasets and optimise for production constraints including model size and inference speed.

Text preprocessing, tokenisation, stemming, and lemmatisation. Bag-of-words, TF-IDF, and word embeddings including Word2Vec and GloVe. Sentiment analysis, text classification, and named entity recognition using classical and deep learning approaches.

The attention mechanism, transformer architecture, and the family of models derived from it — BERT, RoBERTa, DistilBERT, and GPT. Students fine-tune pre-trained transformer models for downstream NLP tasks including question answering, summarisation, and classification.

The architecture of generative models — VAEs, GANs, and diffusion models. Prompt engineering, Retrieval-Augmented Generation, and fine-tuning strategies for large language models. Practical integration of LLM APIs into production applications.

LSTM and GRU architectures for sequential data. Time-series analysis, feature extraction from temporal data, and forecasting applications in finance, energy, and logistics. Students build and evaluate forecasting models on real datasets.

Deploying machine learning models using cloud-managed services: AWS SageMaker, Azure Machine Learning, and Google Vertex AI. Students build end-to-end pipelines from training to serving, including CI/CD for ML models and production monitoring.

Markov decision processes, Q-learning, policy gradients, and the OpenAI Gym environment. Students implement basic reinforcement learning agents and explore applications in robotics, game playing, and sequential decision-making.

Reading, evaluating, and synthesising academic AI research. Students learn to identify genuine advances from hype, reproduce results from published papers, and write technical research summaries — preparation for both university-level study and professional applied research.

The complete lifecycle of a production ML system: continuous integration and deployment for models, automated retraining pipelines, feature stores, A/B testing frameworks, and production monitoring using tools including MLflow, Weights and Biases, and Seldon Core.

Designing AI systems for scale, reliability, and maintainability. Topics include microservices for ML, real-time versus batch inference, data architecture patterns, model serving infrastructure, and the trade-offs between different deployment strategies at enterprise scale.

A comprehensive examination of the regulatory landscape governing AI — the EU AI Act, algorithmic accountability frameworks, and sectoral regulations in healthcare, finance, and criminal justice. Students design an AI governance framework for a simulated organisation.

LIME, SHAP, and attention visualisation techniques that make black-box models interpretable. Students apply explainability methods to medical diagnosis, credit scoring, and hiring models — and evaluate the limitations of each approach in a professional context.

Building sophisticated text generation systems: fine-tuning GPT-class models, instruction tuning, RLHF (Reinforcement Learning from Human Feedback), and evaluation of generated text quality using BLEU, ROUGE, and human assessment protocols.

Graph neural networks for relational data, AutoML for tabular data, and ensemble stacking techniques. Applied to fraud detection, drug discovery, and social network analysis across real-world datasets.

Students select an unsolved problem in their industry of interest and design an original AI solution — from literature review and methodology design through to prototype development and evaluation — supervised by DC Global College academic staff.

Academic writing at postgraduate standard, research proposal design, and the skills required to succeed in a technology MSc abroad. Students complete a full university application including personal statement and research interest statement.