SPECIALIST PROGRAM 02 | THE BRAIN
AI & Machine Learning
Engineering
Don’t just use AI. Build the intelligence behind it.
The Academy’s most technically intensive pathway: machine learning, deep learning, language-model development and agentic AI. Learn to build, train, adapt and evaluate models for enterprise needs, then use them in controlled AI systems.
PROGRAM PURPOSE
Why It Matters
Enterprise AI starts with choosing the right model for the problem. Learn when to use, adapt or train models, including privately hosted approaches for sensitive data, and connect them to applications and agents with human oversight.
Planned Competency Areas
Machine Learning Foundations
Apply Python, mathematics, statistics and core algorithms to regression, classification, clustering, forecasting and anomaly detection.
Model Training & Evaluation
Prepare training data, engineer features, select models, prevent leakage and compare performance through reproducible experiments and error analysis.
Deep Learning Engineering
Build and train neural networks; understand backpropagation, optimisation and regularisation; diagnose failures and improve performance.
LLM Architecture & Training
Understand tokenisation, embeddings, attention and transformers. Build and train a small language model from scratch to learn the principles behind LLMs.
Enterprise Model Adaptation
Fine-tune pretrained models, evaluate retrieval-augmented generation and compare private-hosting options against business, data-control, cost and performance requirements.
Agentic AI Systems
Connect models, tools and workflows through single- and multi-agent systems, with controlled permissions, evaluation, failure handling and human oversight.
Expected Evidence
Trained Machine Learning Model
A model trained for a defined business problem, with baseline comparisons, evaluation results and documented error analysis.
Language Model Built from Scratch
A small language model built and trained from initialisation, with an explanation of its architecture, training process and limitations.
Adapted Enterprise Model
A pretrained model adapted to a business task, with before-and-after evaluation and documented data-handling and hosting choices.
Integrated Agentic System
An application using a trained or adapted model, with tests, failure controls and faculty-reviewed defence of decisions, limitations and improvements.
WITT Relevance
Employer Input
Participating employers may contribute workflow patterns, role requirements, exception types and approved case-study contexts.
Academic Review
Commercial relevance does not override academic independence, fairness, confidentiality or responsible-use requirements.
Candidate Access
Eligible graduates may be considered for relevant WITT opportunities based on evidence, interest, geography and employer criteria.
