Short Course · 4 weeks · Evening
Introduction to AI Systems
A four-week evening course for product managers, analysts, and technical writers who work near AI systems without building them. Covers what a model is and is not, how training data shapes behaviour, where systems fail quietly, and how published governance frameworks describe deployment practice. No coding required.
- Two contact hours weekly
- No assessment — written attendance record issued
- Sources named for every factual claim
- Will not teach: coding, model training, tool walkthroughs
Sources: published AI governance frameworks, course bibliography distributed on day one
Course · 9 weeks · 4 hrs/week + lab
Model Evaluation and Reliability
For engineers who can already train a model and need to know whether it works. Covers test set construction, distribution shift, calibration, ablation design, statistical significance for small samples, adversarial evaluation, documented model limitations, production monitoring, and incident review. Draws on published evaluation literature throughout.
- Four contact hours weekly plus eight hours lab
- Assessment: evaluation suite + written report on model weaknesses
- Participants bring their own model
- Will not teach: model architecture design, deployment engineering
Assessment design draws on Ribeiro et al. (2020) CheckList methodology and Liang et al. (2022) HELM framework
B2B Engagement · Scoped per team
Organisational AI Practice Consulting
A structured engagement for teams introducing AI systems into existing processes. Includes a discovery workshop, written review of current development and review practice, a recommended evaluation and monitoring workflow, three training days for engineering and product staff, a documentation template set, and a review at four months.
- Covers engineering practice, process design, internal capability
- Four-month follow-up review included
- Documentation template set provided
- Does not assess regulatory obligations — those questions go to your own qualified advisers
Workflow recommendations drawn from published frameworks; regulatory compliance questions are outside scope
