THE COMPLETE ROUTE

Learning Journey

Welcome to the Course Rider. Land, fasten your seat belt and ride eleven milestones from orientation to capstone — with pit stops, focus checks and a memory verse for every stage.

Pit stop: Step away for ten minutes. Recall beats rereading.

Focus radar: Name three things you can now explain out loud.

START HERE

Orientation

Before Day 1 · 45 min

Understand how this programme connects Data, Models, Machine Learning, Deep Learning, Transformers, LLMs, Generative AI, RAG and Agents into one complete enterprise AI system.

Memory verse: Know the route before the run.

TOPICS COVERED

  • Programme overview and learning outcomes
  • Required tools: Python, Jupyter, VS Code
  • Student baseline self-assessment
  • Project Studio introduction
  • How assessment, badge and certificate work

LEARNING OUTCOME

Know the route, the tools and the project you will build across twenty days.

PROJECT STUDIO STEP

Choose the enterprise scenario your Project Studio build will follow for the whole programme.

Memory verses

Memory verses

One short line per milestone — small enough to remember today, strong enough to explain the whole concept tomorrow. Flip a card for the motivation behind it.

1 / 11

Advanced concept path

ADVANCED CONCEPT PATH

From raw data to autonomous action

Six layers, each consuming the output of the one before it. Select any node to see what it means and where it sits in the chain.

FOUNDATION LAYERThe business problem and the governed data that describes it.
MODEL LAYERTurning prepared data into an evaluated, servable artifact.
DEEP LEARNING LAYERLayered networks that learn their own representations.
LANGUAGE INTELLIGENCE LAYERHow text becomes vectors, attention and generated tokens.
KNOWLEDGE LAYERGrounding generation in evidence you own and control.
ACTION LAYERFrom an answer to an approved, audited business action.

FOUNDATION LAYER

Business problem

The decision the system must improve, with a measurable target.

END-TO-END CHAIN

Business ProblemDataFeaturesAlgorithmModelPredictionTransformerLLMGenerative AIEmbeddingsRAGAgentToolsHuman ApprovalBusiness Outcome

Lifecycles you will run

Every stage lists its inputs, activities, outputs, tools, the common mistake and an illustrative industry case.

PROCESS FLOW

AI system lifecycle

From a business problem to a monitored, governed system.

  1. A framing step converts a vague ambition into a testable objective: who decides, how often, what a good outcome looks like and what the cost of an error is. Without it, every later technical choice is unfalsifiable.

    INPUTS

    • Stakeholder interviews
    • Current process description
    • Baseline performance

    ACTIVITIES

    • Define the decision
    • Agree a success metric
    • Estimate error cost

    OUTPUTS

    • Problem statement
    • Success metric
    • Scope boundary

    TOOLS

    • Workshops
    • Decision records

    COMMON MISTAKE · Building a model for a decision nobody actually makes.

    INDUSTRY · Hypothetical: a telecom operator frames churn as 'which accounts to call this week', not 'predict churn'.

PROCESS FLOW

RAG implementation lifecycle

Indexing private documents, then answering with cited evidence.

  1. Parsers extract text, headings and tables while preserving the source path, owner, date and permission label so retrieved evidence can always be attributed and filtered.

    INPUTS

    • PDFs, documents, pages
    • Permission model

    ACTIVITIES

    • Extract text
    • Preserve structure
    • Attach metadata

    OUTPUTS

    • Normalised documents

    TOOLS

    • Document parsers
    • OCR

    COMMON MISTAKE · Losing headings so chunks become context-free fragments.

    INDUSTRY · Hypothetical: an HR team indexes policy PDFs with department and effective-date metadata.

PROCESS FLOW

Agent workflow lifecycle

Goal to business action, with a human in control.

  1. Goals are scoped in advance. An agent with an open-ended objective and broad credentials is an operational risk, so the goal is expressed as a task type with explicit limits.

    INPUTS

    • User request
    • Policy scope

    ACTIVITIES

    • Parse intent
    • Check scope

    OUTPUTS

    • Bounded goal

    TOOLS

    • Intent classification

    COMMON MISTAKE · Accepting goals outside the approved task catalogue.

    INDUSTRY · Hypothetical: a delivery team allows ticket creation but not customer refunds.

Every domain in three phases

Fundamentals, Foundations and Application — the same academic ladder applied to each subject in the programme.

THREE-PHASE FRAMEWORK

Artificial Intelligence

What AI names, how AI systems are assembled, and how they are governed in production.

  1. PHASE 1

    Fundamentals

    What intelligence and Artificial Intelligence mean, and why the field exists.

    • Intelligence and perception
    • AI versus automation
    • AI versus analytics
    • Narrow AI, AGI, ASI
  2. PHASE 2

    Foundations

    The families of techniques and the anatomy of a working AI system.

    • Symbolic and statistical AI
    • Machine Learning and Deep Learning
    • Data, model, inference, feedback
    • AI system lifecycle
  3. PHASE 3

    Application

    How AI is delivered, evaluated, governed and trusted in an enterprise.

    • Use-case selection
    • Evaluation and monitoring
    • Human oversight
    • Responsible AI policy

THREE-PHASE FRAMEWORK

Data

From raw values to governed, retrievable enterprise knowledge.

  1. PHASE 1

    Fundamentals

    The vocabulary of data before any tooling.

    • Data, information, knowledge
    • Values, fields, records
    • Datasets and schemas
    • Measurement scales
  2. PHASE 2

    Foundations

    Classification, quality and where data lives.

    • Structured to unstructured
    • Quality dimensions
    • Lifecycle and governance
    • Databases, warehouses, lakes, vector stores
  3. PHASE 3

    Application

    Turning data into decisions and grounded answers.

    • Cleaning and feature engineering
    • Analysis and reporting
    • Machine Learning inputs
    • RAG corpora and business decisions

THREE-PHASE FRAMEWORK

Models

A model is a learned function with a lifecycle, not a file.

  1. PHASE 1

    Fundamentals

    What a model is and what it consumes.

    • Inputs, processing, output
    • Features and labels
    • Prediction versus rule
    • Parameters versus hyperparameters
  2. PHASE 2

    Foundations

    How a model is built honestly.

    • Train, validation, test
    • Leakage and baselines
    • Generalisation
    • Model selection
  3. PHASE 3

    Application

    Serving and maintaining the artifact.

    • Packaging and registries
    • Inference APIs
    • Monitoring and drift
    • Retraining policy

THREE-PHASE FRAMEWORK

Machine Learning

Learning patterns from data, then defending the result with evidence.

  1. PHASE 1

    Fundamentals

    Learning from data instead of hand-written rules.

    • Supervised learning
    • Unsupervised learning
    • Reinforcement learning
    • Task types
  2. PHASE 2

    Foundations

    The mechanics behind a trained model.

    • Features and labels
    • Algorithms and training
    • Evaluation metrics
    • Bias and variance
  3. PHASE 3

    Application

    Machine Learning inside a business system.

    • Pipelines and APIs
    • Deployment
    • Monitoring and drift
    • Retraining and business integration

THREE-PHASE FRAMEWORK

Deep Learning

Layered representation learning, from a single neuron to a trained network.

  1. PHASE 1

    Fundamentals

    The artificial neuron.

    • Inputs and weights
    • Bias
    • Activation functions
    • Why depth helps
  2. PHASE 2

    Foundations

    How a network actually learns.

    • Layers and forward propagation
    • Loss functions
    • Backpropagation
    • Gradient descent
  3. PHASE 3

    Application

    Architectures and frameworks in practice.

    • CNNs for vision
    • RNNs and LSTMs for sequence
    • Regularisation
    • TensorFlow, Keras, PyTorch

THREE-PHASE FRAMEWORK

Transformers

Attention replaced recurrence and made modern language models possible.

  1. PHASE 1

    Fundamentals

    Turning language into numbers.

    • Tokenisation
    • Token IDs
    • Embeddings
    • Positional information
  2. PHASE 2

    Foundations

    The attention mechanism.

    • Query, Key, Value
    • Self-attention
    • Multi-head attention
    • Transformer block
  3. PHASE 3

    Application

    Encoders, decoders and downstream use.

    • Encoder versus decoder stacks
    • Pretraining objectives
    • Context windows
    • Serving considerations

THREE-PHASE FRAMEWORK

Large Language Models

Next-token prediction at scale — powerful, and confidently wrong when ungrounded.

  1. PHASE 1

    Fundamentals

    What an LLM is and is not.

    • Next-token prediction
    • Statistical generation, not reasoning like a person
    • Prompt and completion
    • Context window
  2. PHASE 2

    Foundations

    How outputs are shaped.

    • Pretraining and alignment
    • Sampling: temperature and top-p
    • System and user roles
    • Hallucination causes
  3. PHASE 3

    Application

    Building on an LLM responsibly.

    • Grounding and citations
    • Evaluation harnesses
    • Cost and latency
    • Safety and logging

THREE-PHASE FRAMEWORK

Generative AI

Producing new artefacts under constraint, with review built in.

  1. PHASE 1

    Fundamentals

    Prediction versus generation.

    • Discriminative versus generative
    • Modalities
    • Prompt anatomy
    • Determinism limits
  2. PHASE 2

    Foundations

    Controlling generation.

    • Structured output
    • Prompt chaining
    • Few-shot patterns
    • Guardrails
  3. PHASE 3

    Application

    Generation inside a product.

    • Templates and brand rules
    • Human review
    • Provenance
    • Cost control

THREE-PHASE FRAMEWORK

RAG

Retrieval-Augmented Generation grounds answers in evidence you control.

  1. PHASE 1

    Fundamentals

    Why retrieval exists.

    • Model knowledge cut-off
    • Private documents
    • Embeddings
    • Similarity
  2. PHASE 2

    Foundations

    The indexing and query pipelines.

    • Parsing and chunking
    • Metadata
    • Vector search
    • Reranking
  3. PHASE 3

    Application

    Trustworthy grounded answers.

    • Context assembly
    • Citations
    • Permission filtering
    • RAG evaluation

THREE-PHASE FRAMEWORK

Agents

Systems that plan and act — inside explicit permission boundaries.

  1. PHASE 1

    Fundamentals

    Model versus workflow versus agent.

    • Goal
    • State
    • Tool
    • Autonomy levels
  2. PHASE 2

    Foundations

    The agent loop.

    • Planning
    • Tool selection and function calling
    • Memory
    • Verification and retry
  3. PHASE 3

    Application

    Agents an enterprise can approve.

    • Human approval gates
    • Scoped credentials
    • Observability
    • Audit logs

THREE-PHASE FRAMEWORK

Responsible AI

Not a final chapter — a constraint applied at every layer.

  1. PHASE 1

    Fundamentals

    The obligations.

    • Privacy
    • Fairness
    • Transparency
    • Accountability
  2. PHASE 2

    Foundations

    Making obligations measurable.

    • Data minimisation
    • Bias testing
    • Explainability
    • Source validation
  3. PHASE 3

    Application

    Operating safely.

    • Human oversight
    • Incident response
    • Model and data documentation
    • Regulatory alignment

THREE-PHASE FRAMEWORK

Enterprise AI Architecture

How every layer of this programme fits into one production estate.

  1. PHASE 1

    Fundamentals

    The layers.

    • Data layer
    • Model layer
    • Knowledge layer
    • Action layer
  2. PHASE 2

    Foundations

    The connective tissue.

    • Pipelines and orchestration
    • Registries and feature stores
    • Identity and permissions
    • Observability
  3. PHASE 3

    Application

    Running it.

    • MLOps and LLMOps
    • Cost governance
    • Change management
    • Continuous improvement