AI Solutions Architecture

The course we teach for working architects: from the AI/ML landscape and MLOps to generative AI, agents, explainability, cloud platforms and designing solutions that hold up in production.

Practitioner17 sessions · 90 minutes each17 lessons

Taught live over eight weeks, sixteen lessons plus a welcome session. Each lesson pairs concepts with a demonstration or a structured thought experiment, and the course closes with a full solution-architecture case study.

Materials here are members-only.

00Welcome and orientationWhat the course is, and is not, about. Baseline definitions of AI, machine learning, deep learning, generative AI and foundation models.Interactive: Where the terms sitFree · read now
01The AI/ML landscapeThe problems AI solves, a typical project lifecycle and the roles involved. The AI/ML architect compared with the generative AI architect.Interactive: A project, end to endMembers
02Machine learning operationsWhy "launch a model" is not enough. The MLOps pipeline from data to retraining, and what GenAIOps adds: prompt catalogues, prompt flows, model gardens and agents.Interactive: The MLOps pipelineMembers
03Data and data engineeringAggregation, ETL and preparation; feature engineering, imputation and cleansing, with Pandas, PySpark and Ray.Interactive: Cleaning a datasetMembers
04Machine learning I: regressionSupervised and unsupervised learning. Linear regression, gradient-boosted trees and decision trees.Interactive: Fitting a regressionMembers
05Machine learning II: classification and clusteringLogistic regression, k-means, principal component analysis and autoencoders, with a decision map for choosing a method.Interactive: k-means, one step at a time · Principal componentsMembers
06Training, tuning and deploymentHyperparameter tuning, deployment patterns from real-time to batch, blue/green releases, monitoring, drift and retraining loops.Interactive: Gradient descent and the learning rate · Drift and retrainingMembers
07Deep learning and reinforcement learningNetwork essentials, building a model in Keras or PyTorch, and the components of reinforcement learning.Interactive: A neural network learning · Reinforcement learning in a gridMembers
08Advanced deep learningLanguage understanding with tokenisation and embeddings, modern time-series forecasting, and computer vision with convolutional networks.Interactive: Attention, token by token · A convolution, step by step · Forecasting and anomaliesMembers
09Generative AI I: conceptsHow transformers, diffusion and reasoning models work, and how large language models behave in practice.Interactive: Choosing the next token · Diffusion: from noise to imageMembers
10Generative AI II: prompts and multi-agent systemsPrompt engineering fundamentals and an overview of multi-agent systems, with a chatbot demonstration.Interactive: Multi-agent patternsMembers
11Visualisation and applicationsCommunicating model results and building AI applications, with a Streamlit dashboard demonstration.Interactive: One dataset, four picturesMembers
12AI at scaleLarge-scale and distributed training, and managing models across data and architecture.Interactive: Training across many acceleratorsMembers
13The state of the artReasoning models, hardware acceleration, small language models and what lies beyond today's agents.Interactive: Spending compute at answer timeMembers
14Explainable AIExplainability and interpretability, the bias-variance trade-off, SHAP, LIME, fairness assessment and ethics.Interactive: Explaining one prediction · Bias and varianceMembers
15Cloud services for AIThe major platforms compared: SageMaker and Bedrock, Azure AI Foundry, Vertex AI, watsonx and the wider ecosystem.Interactive: Who runs which layerMembers
16Becoming an AI architectFrom business need to requirements and a well-architected design: resilience, security, sustainability and cost. Portfolio and interview practice.Interactive: Well-architected trade-offsMembers