Curriculum
Each day is a full textbook module with learning objectives, diagrams, industry examples, Python labs, quizzes and an assignment.
- Day 1Fundamental · 75 min
Fundamentals of Artificial Intelligence
Establish a precise working definition of intelligence and artificial intelligence, separate AI from automation and analytics, and learn to read any AI system as an input, processing and output pipeline.
CONCEPTS
IntelligenceArtificial IntelligencePerception and reasoningAutomation vs AIAnalytics vs AIInput → Model → OutputLab: Identify AI, automation and analytics in ten real-world scenarios.
Unlocks after enrolment
- Day 2Foundation · 80 min
Foundations of Artificial Intelligence
Trace AI from symbolic reasoning to foundation models, classify system types from reactive to self-aware, and learn the anatomy and lifecycle of a complete AI system including its responsible-AI obligations.
CONCEPTS
Symbolic AIRule-based systemsSearch and planningStatistical AIDeep learningFoundation modelsLab: Map the components of a real AI application.
Unlocks after enrolment
- Day 3Fundamental · 70 min
Fundamentals of Data
Define data, information and knowledge precisely, follow the DIKA chain to action, and classify any field by nature and measurement scale — the skill every later module depends on.
CONCEPTS
DataInformationKnowledgeMetadataQualitative vs quantitativeNominal and ordinalLab: Classify twenty fields by nature and measurement scale.
Unlocks after enrolment
- Day 4Foundation · 85 min
Foundations of Data
Structure, source, time and sensitivity classifications; the six dimensions of data quality; the data lifecycle and governance; and how databases, warehouses, lakes, lakehouses and vector stores differ.
CONCEPTS
Structured / semi-structured / unstructuredPrimary and secondary dataStreaming and time-seriesSensitivity classificationSix quality dimensionsData lifecycleLab: Profile, clean and classify a customer dataset.
Unlocks after enrolment
- Day 5Fundamental · 90 min
Python and Data Tools for AI
The working Python subset used in AI engineering: core types and control flow, files, CSV and JSON, NumPy, Pandas, Matplotlib, APIs, notebooks, virtual environments and safe handling of secrets.
CONCEPTS
Types and collectionsControl flow and functionsCSV and JSONNumPy vectorisationPandas DataFrameMatplotlibLab: Load a CSV with Pandas, inspect it, clean it, calculate KPIs and create a chart.
Unlocks after enrolment
- Day 6Fundamental · 70 min
Fundamentals of Models
What a model actually is: a learned mathematical function with parameters. Features, targets, weights, bias, hyperparameters, the algorithm/model distinction and the training/inference distinction.
CONCEPTS
ModelFeatureTargetWeightBiasHyperparameterLab: Build a simple linear model and inspect its inputs, parameters and outputs.
Unlocks after enrolment
- Day 7Foundation · 85 min
Foundations of Model Building
The disciplined lifecycle of producing a model: problem framing, success criteria, splits and cross-validation, data leakage, baselines, selection, tuning, deployment, monitoring, retraining and versioning.
CONCEPTS
Problem framingSuccess criteriaTrain/validation/testData leakageCross-validationBaselineLab: Design a complete model lifecycle for customer churn.
Unlocks after enrolment
- Day 8Fundamental · 75 min
Fundamentals of Machine Learning
Learning paradigms and task types: supervised, unsupervised and reinforcement learning; classification, regression, clustering, anomaly detection and recommendation; generalisation and confidence.
CONCEPTS
SupervisedUnsupervisedReinforcementClassificationRegressionClusteringLab: Match different business problems with the correct machine learning approach.
Unlocks after enrolment
- Day 9Intermediate · 100 min
Machine Learning Algorithms and Evaluation
The working algorithm set — linear and logistic regression, trees, random forests, gradient boosting, SVM, kNN, k-means, PCA — with the confusion matrix, precision, recall, F1, ROC-AUC, error metrics, imbalance, bias–variance and explainability.
CONCEPTS
Linear/logistic regressionDecision treesRandom forestGradient boostingSVMkNNLab: Train and compare Logistic Regression, Random Forest and XGBoost on a churn dataset.
Unlocks after enrolment
- Day 10Intermediate · 90 min
Machine Learning in Production
Turning a validated model into a dependable service: pipelines, serialisation, batch and real-time inference, FastAPI and Streamlit, registries, containerisation, drift detection, retraining, rollback and human review.
CONCEPTS
PipelinesJoblibBatch vs real-timeFastAPIStreamlitModel registryLab: Deploy a churn model through a Streamlit interface.
Unlocks after enrolment
- Day 11Fundamental · 85 min
Fundamentals of Deep Learning
The artificial neuron in full detail: inputs, weights, bias, weighted sum, activation, layers, forward propagation, loss, epochs, batches and learning rate — computed by hand and in code.
CONCEPTS
Artificial neuronWeights and biasActivationLayersForward propagationLossLab: Calculate the output of a small artificial neuron manually and in Python.
Unlocks after enrolment
- Day 12Foundation · 95 min
Foundations of Deep Learning
How networks actually learn: backpropagation and gradient descent, optimisers, activation and loss selection, CNNs, RNNs, LSTMs and GRUs, dropout, batch normalisation, regularisation, early stopping and the frameworks.
CONCEPTS
BackpropagationGradient descentOptimisersReLU/sigmoid/softmax/tanhCross-entropyCNNLab: Build a Keras neural network and visualise training and validation accuracy.
Unlocks after enrolment
- Day 13Deep Dive · 105 min
Transformers and Attention
Tokenisation, embeddings, positional information, query/key/value self-attention, scaled dot-product, multi-head attention, feed-forward blocks, residuals, normalisation, and encoder/decoder variants.
CONCEPTS
TokenisationToken IDsEmbeddingsPositional encodingSelf-attentionQ/K/VLab: Use an interactive sentence to demonstrate how one word attends to other relevant words.
Unlocks after enrolment
- Day 14Deep Dive · 100 min
Large Language Models
Pretraining and next-token prediction, context windows, the generation loop, temperature and top-k/top-p sampling, system and user prompts, instruction tuning, preference optimisation, hallucination and knowledge cutoff.
CONCEPTS
PretrainingNext-token predictionContext windowTemperatureTop-kTop-pLab: Compare responses using different temperatures and prompt structures.
Unlocks after enrolment
- Day 15Intermediate · 90 min
Generative AI
Predictive versus generative AI, generation across modalities, prompt engineering as a specifiable discipline, structured output, prompt chaining, grounding, hallucination control, evaluation and human review.
CONCEPTS
Predictive vs generativeMultimodalPrompt engineeringZero-shotFew-shotStructured outputLab: Build a prompt that converts raw project status data into a structured executive summary.
Unlocks after enrolment
- Day 16Intermediate · 90 min
Adapting Models: Prompting, RAG, Fine-Tuning and LoRA
How to change a model's behaviour without pretraining it: the four adaptation levers, what each one can and cannot fix, dataset design for supervised fine-tuning, parameter-efficient methods such as LoRA, and the cost and governance consequences of each choice.
CONCEPTS
Adaptation ladderSupervised fine-tuningInstruction dataLoRA and adaptersQuantisationCatastrophic forgettingLab: Design a 50-example instruction dataset and run a LoRA fine-tune configuration review.
Unlocks after enrolment
- Day 17Intermediate · 85 min
Embeddings and Vector Search
Meaning as geometry: how text becomes vectors, why cosine similarity works, how chunking decides retrieval quality, what an approximate nearest-neighbour index actually trades away, and how to evaluate a search system with recall@k and MRR.
CONCEPTS
EmbeddingVector spaceCosine similarityChunkingANN indexHybrid searchLab: Build a small semantic search index over a document set and measure recall@5.
Unlocks after enrolment
- Day 18Deep Dive · 90 min
Retrieval-Augmented Generation
Assembling retrieval and generation into a system that answers from evidence: the full RAG pipeline, reranking, context assembly, citation enforcement, refusal behaviour, faithfulness evaluation and the failure modes that make RAG systems quietly wrong.
CONCEPTS
RAG pipelineQuery rewritingRerankingContext assemblyCitationsRefusalLab: Build a cited question-answering pipeline with a refusal path and score its faithfulness.
Unlocks after enrolment
- Day 19Deep Dive · 90 min
AI Agents: Tools, Memory and Planning
What changes when a model can act: the agent loop, tool schemas and validation, memory tiers, planning strategies, termination conditions, permission design and human approval for consequential actions.
CONCEPTS
Agent loopTool callingFunction schemaMemory tiersPlanningReflectionLab: Build a bounded two-tool agent with schema validation, budgets and an approval gate.
Unlocks after enrolment
- Day 20Deep Dive · 95 min
Enterprise AI: Architecture, Observability and Governance
The capstone: a reference architecture for AI in an organisation, the gateway pattern, cost and latency engineering, observability and tracing, continuous evaluation, incident response, governance under real regulation, and the operating model that sustains it.
CONCEPTS
Reference architectureAI gatewaySemantic cachingModel routingTracingContinuous evaluationLab: Design an end-to-end architecture and observability plan for a regulated AI assistant.
Unlocks after enrolment