COMP328 Fundamentals of Machine Learning

Module 6 of 6 · Final module

Responsible and Applied Machine Learning

The final module shifts from building models to deploying them responsibly. You will learn what happens to a model after training — how it reaches users, how it can fail, who it can harm, and how regulation is beginning to shape what practitioners are permitted to build. The module closes with a structured integration of the entire course, in preparation for the capstone.

4 units 36 tutorials 4 unit quizzes 1 capstone project

Module at a glance

Units 4 — deployment, ethics, governance, integration
Tutorials 36 — nine per unit
Quizzes 4 — one at the end of each unit
Capstone Released in Unit 20, due week 14
Study time ~22 hours plus capstone work

Learning outcomes

What you will be able to do

By the end of Module 6 you will be able to take a model beyond the notebook — deploy it, monitor it, assess its ethical implications, and reason about the regulatory environment it operates within.

By the end of this module you will be able to

  • Describe the stages of a production machine learning pipeline, from data ingestion through to serving and monitoring.
  • Explain how a trained model is packaged and served through an API, and identify the practical differences between batch and real-time inference.
  • Recognise the ways a deployed model can degrade — data drift, concept drift, and feedback loops — and describe monitoring strategies for each.
  • Identify sources of bias in machine learning systems, including historical bias, representation bias, and measurement bias, and distinguish them from statistical variance.
  • Analyse the privacy implications of collecting and modelling personal data, and describe techniques such as anonymisation, differential privacy, and federated learning.
  • Explain the core principles of algorithmic fairness, including the tension between different formal definitions of fairness and why no single definition suffices.
  • Assess a machine learning system against the emerging regulatory landscape, including data protection law, sector-specific regulation, and proposed AI-specific frameworks.
  • Integrate technical, ethical, and professional considerations into a coherent argument about whether and how a machine learning system should be deployed.

The final module is about judgement

Modules 1 through 5 taught you how to build models. Module 6 asks whether you should. The technical content here is lighter than Module 5, but the reasoning is denser — you will be asked to form and defend positions on questions that do not have clean technical answers.

Prerequisite reminder

This module assumes you have completed Modules 1 through 5, or that you have equivalent experience. The ethics and governance discussions refer to techniques from every earlier module — regression, classification, clustering, and neural networks all appear as examples. Unit 20 in particular assumes working familiarity with the full course.

Units & tutorials

Four units, thirty-six tutorials

Units 17, 18, and 19 are substantive and independent — deployment, ethics, and governance respectively. Unit 20 ties the entire course together and launches the capstone.

Unit 17 9 tutorials · ~6 hours

Model Deployment

What happens to a model after the notebook is closed. Production pipelines, APIs, monitoring, and the emerging discipline of MLOps.

Unit overview Start T-01
Unit 18 9 tutorials · ~6 hours

Ethics

Machine learning systems make decisions about people. This unit examines how they can cause harm, and what practitioners owe to those affected by their work.

Unit overview Start T-01
Unit 19 9 tutorials · ~6 hours

Governance and Fairness

Ethics as a set of principles is not enough. This unit examines the formal definitions of fairness, the regulatory landscape, and the institutional structures that are emerging to govern AI systems.

Unit overview Start T-01
Unit 20 9 tutorials · ~4 hours

Course Integration

The final unit. A structured review of the entire course, a preparation guide for the capstone, and a look at where the field is going next.

Unit overview Start T-01

Capstone

The final project

The capstone is a complete machine learning project that integrates everything from the course. It is released during Unit 20 and due in week 14, immediately before the final exam period.

Design, build, evaluate, and defend

You will select a problem, acquire and clean data, build and evaluate a model, interpret the results, analyse the ethical implications, and present your work. Every stage is assessed, and every stage draws on an earlier module.

Capstone overview Instructions
Problem Selection Define a question worth asking and answerable with data.
Data Acquisition & Cleaning Source, inspect, and prepare a dataset for modelling.
Model Building & Evaluation Train, compare, and select a model using appropriate metrics.
Interpretation & Ethics Explain what the model says, and what it might get wrong.
Final Report A written document covering the full lifecycle.
Presentation A spoken defence of your work to the course cohort.

How to work through it

Read actively, argue carefully, then build

Module 6 is the most discursive module in the course. The habits that matter here are different again — closer to a seminar than a laboratory.

  1. Read the case studies closely

    The ethics unit is built around real systems that caused real harm. Read each case study slowly and try to identify exactly where the failure originated.

  2. Form positions and defend them

    The unit quizzes in Module 6 ask for reasoning, not just recall. When a tutorial presents a trade-off, decide where you stand before reading the summary.

  3. Apply everything to your capstone

    As you work through Units 17 to 19, keep asking how each idea applies to the project you are planning. The capstone will ask you to do exactly that.

  4. Finish with the unit quiz

    Each unit ends with a short quiz. In this module, the quizzes weight reasoning and application more heavily than earlier ones.

Assessment

How this module is assessed

Module 6 carries four unit quizzes and launches the capstone project, which accounts for 40% of your final grade.

  • Quiz 17 Model Deployment — 15 questions, counts toward your quiz grade Unit 17
  • Quiz 18 Ethics — 15 questions, reasoning-weighted Unit 18
  • Quiz 19 Governance and Fairness — 15 questions, reasoning-weighted Unit 19
  • Quiz 20 Course Integration — 20 questions, comprehensive review Unit 20
  • Capstone End-to-end machine learning project — 40% of final grade Open

Capstone deadlines are strict

The capstone is released during Unit 20 and due in week 14. Because it directly precedes the final exam period, there is no extension window. Submit your problem selection and dataset choice early — the instructions describe a two-stage submission process with an interim checkpoint.

Quiz 20 is a comprehensive review

Unlike the other module quizzes, Quiz 20 draws on material from all six modules. It is designed to be diagnostic rather than punitive — treat your score as a map of what to revisit before the capstone defence. Review the course schedule to see how it fits into the final weeks.

👤