Applied AI Engineering

Empowers individuals to scope, prototype and evidence “quick-win” ML services designed with commercial KPIs in mind. Our learners start looking at worked examples of Mathematics in the workplace, used to optimise performance metrics in the business context.

Workshop 2: Collaborative Engineering With AI: A Rapid MVP Delivery Pipeline Using Public Models

Task: Re-use a containerised Proof-of-Concept ML model to create a prototype solving a real-world business problem, produce an ROI snapshot for stakeholders; pitch the defined saving or uplift

A hands-on dive into Coding with Python, version control and core ML libraries. Emphasis is placed on clean code, test-driven development and reusable components that underpin production AI systems. Explores parallelism, algorithmic thinking and best programming practices.

Workshop 3: Pair-programming & debugging clinic for fun and profit

Task: Refactor legacy code into a reusable, unit-tested ML module and simulate interaction with a Software Engineer

This unit covers data sourcing, cloud pipelines, secure cloud storage and cybersecurity best practices, linking technical design to cost, scalability and compliance. Creation of architecture diagrams that balance performance with governance. Booster modules brush up on networking and other relevant knowledge.

Workshop 4: Architecting secure pipelines with IaC (Infrastructure-as-Code)

Task: Produce a model for a proposed data flow and risk it

Master CI/CD, containerisation and live-model monitoring. The focus is on deploying robust pipelines and instituting incident-response procedures that keep AI services healthy. We will focus on model lifecycle management, staging, experiment tracking and live monitoring.

Workshop 5: End-to-end AI/ML deployment and orchestration

Task: Create an MLFlow-tracked workflow. Ship an orchestrated model to a staging environment and set alert thresholds for service management

Builds understanding of vectors, matrices, statistics and optimisation in a practical, code-first knowledge/skills/behaviours unit. Connects mathematical theory to real ML tuning problems and performance metrics.

Workshop 6: Metrics and model-tuning lab (metric-driven optimisation)

Task: Compare optimiser performance on a real dataset and simulate collaboration with a data scientist

Regression, classification and evaluation techniques take centre stage. Bias-variance trade-offs and model-selection strategies prepare participants for production scenarios. We also cover computer vision and evaluation strategies.

Workshop 7: Classification in action: production-grade classifiers

Task: Train, evaluate and document a supervised model for a business dataset. Containerise it and publish an API with contracted tests and version tags

From data cleaning to feature extraction and stability testing, this unit shows how strong features make or break model performance. We teach robust feature pipelines, drift checks and automated unit tests.

Workshop 8: Influencing model success with model drift and data drift safeguard builds

Task: Build a feature pipeline with automated checks. Create a feature store script with built-in monitoring and push it to the registry

Mid-programme hackathon team challenge to deliver an end-to-end ML solution onto a sandbox cluster.

Clustering, semi-supervised, ensemble and reinforcement learning are explored here, with an eye on when each approach delivers business value.

Workshop 9: Putting it all together: model selection play-off

Task: Compare three unsupervised algorithms on the same dataset and justify the preferred choice. Publish a comparative dashboard and recommend a production candidate

Design, train and debug neural networks, weighing GPU cost against accuracy. In this unit, popular architectures such as CNNs, RNNs and transformers, are demystified.

Workshop 10: Build, train and benchmark a neural net

Task: Tune hyper-parameters to beat a published baseline on a structured dataset

Transfer learning, fine-tuning and pruning illustrates how to squeeze extra performance from models while keeping costs down.

Workshop 11: Optimising a pre-trained model with Hugging Face

Task: Adapt a public model to a novel dataset and log the improvement curve

Focus moves to NLP and domain-specific AI solutions. Aligns technical deliverables with strategic outcomes, commercial goals and stakeholder needs.

Workshop 12: Sentiment analysis for commercial insight

Task: Build a text analytics service and integrate it with a Business Intelligence platform to produce a business impact memo informing strategy

Retrieval-augmented generation, prompt-engineering and scalable LLM hosting are covered, highlighting opportunities and risks of GenAI, covering RAG patterns and secure GenAI APIs operating in agentic architectures.

Workshop 13: Hands-on GenAI deployment: rolling out an agentic LLM service

Task: Deploy a small API-enabled RAG-powered service and quantify the most relevant metrics for optimisation

Fairness, visualisation and explainability (XAI) techniques facilitates communication of model behaviour to technical and non-technical audiences.

Workshop 14: Explainability dashboard sprint and user-centric critical analysis

Task: Produce a SHAP-based report and accompanying stakeholder slide deck

Scoping the EPA project. Practises stakeholder engagement and map career pathways, integrating all previous skills. We focus on integrating vs. decommissioning legacy systems, horizon scanning with technology scorecards, managing innovation with roadmaps and honing research skills.

Workshop 15: Project strategy & stakeholder masterclass

Task: Finalise capstone documentation and evaluation

This final unit provides preparation for successful transition into the End Point Assessment (EPA) period. Participants will consolidate their learning, reflect on their progress, and refine the evidence and artefacts required for assessment. Through guided preparation this unit equips individuals to complete their EPA confidently, but to take the next step in their ongoing professional development.