Capstone: AI Engineering Project

About this course

The engineering capstone for the AI Engineer and Generative AI Engineer pathways. Each learner (or a pair) takes a real problem from framing to a deployed, evaluated and documented ML or LLM-based system in eight weeks. Weekly milestone sessions act as design and progress reviews, mentors meet learners one to one, and the project ends with a live demo and defence before a panel that includes an industry guest.

For: student, software engineer, data professional, ml practitioner.

Learning outcomes

On successful completion you will be able to:

1. 1Frame a real problem as an AI engineering project with stakeholders, success metrics, constraints, risks and a data plan.
2. 2Design and build a working ML or LLM-based system to production standard, with tests, continuous integration and an automated deployment.
3. 3Evaluate the system with an offline evaluation set and user or online evidence, and report results and limitations honestly.
4. 4Apply security, privacy and responsible-AI controls appropriate to the system's risk, and document them.
5. 5Demonstrate comprehensive knowledge of the models, data and infrastructure the system relies on, including their failure modes and the alternatives considered.
6. 6Manage the project autonomously to a fixed timeline, act on mentor and peer feedback, and present and defend the work to a review panel.

Modules and lessons

1. Week 1Project charter⌄

   • Project charter and problem framing
2. Week 2Requirements and evaluation plan⌄

   • Requirements, quality targets and the evaluation plan
3. Week 3System design review⌄

   • System design review
4. Week 4First end-to-end slice⌄

   • Build check-in: the first end-to-end slice
5. Week 5Evaluation and testing review⌄

   • Evaluation and testing review
6. Week 6Security and responsible-AI review⌄

   • Security, privacy and responsible-AI review
7. Week 7Final defence⌄

   • Final demo and panel defence
8. Week 8Retrospective and portfolio⌄

   • Retrospective and portfolio publication

Assessment

• Project (50%)

  The working system - repository, deployment, tests, README, ADRs and a security and responsible-AI note.

• Design review (20%)

  Evaluation report and technical review at the week-5 milestone.

• Viva (30%)

  Final live demo and thirty-minute defence before two mentors and an industry guest.

Pass mark 60%. Grades: Distinction, Merit, Pass.