Deploy, Evaluate, and Create AI Systems Course

Lesson 1 • AI Use Cases Across Industries

Surveys real-world AI applications in business, healthcare, and operations. Grounds abstract concepts in concrete deployment scenarios students will encounter.

Lesson 2 • Data Fundamentals for AI

Covers data types, quality dimensions, and preprocessing basics essential for AI. Prepares students to assess data readiness before model development begins.

Lesson 3 • AI Concepts and Terminology

Defines machine learning, deep learning, and generative AI distinctions. Provides shared vocabulary used throughout all subsequent chapters.

Lesson 4 • Anatomy of an AI System

Maps the components of a production AI system from data ingestion to output delivery. Connects component roles to deployment and evaluation responsibilities.

Lesson 1 • Data Sourcing and Collection

Examines primary and secondary data sources and collection strategies. Establishes data provenance practices critical for reproducibility and compliance.

Lesson 2 • Bias Detection and Mitigation in Data

Identifies sources of bias in datasets and applies mitigation techniques. Ensures data pipelines produce fair inputs before model training begins.

Lesson 3 • Data Cleaning and Preprocessing

Teaches techniques for handling missing values, outliers, and inconsistencies. Clean data directly reduces model error and downstream evaluation noise.

Lesson 4 • Building Scalable Data Pipelines

Designs automated, repeatable pipelines for continuous data delivery to models. Scalable pipelines reduce manual effort and support production-grade AI systems.

Lesson 1 • Transfer Learning and Fine-Tuning

Uses pretrained models to accelerate development on new tasks. Reduces data and compute requirements while achieving competitive performance.

Lesson 2 • Choosing the Right Model Architecture

Compares model families and maps them to task types and data constraints. Correct architecture selection prevents costly retraining cycles later.

Lesson 3 • Experiment Tracking and Reproducibility

Implements logging, versioning, and artifact management for training runs. Reproducibility is a prerequisite for reliable evaluation and team collaboration.

Lesson 4 • Hyperparameter Tuning

Applies systematic search strategies to optimise model hyperparameters. Tuning directly improves generalisation and reduces overfitting in production models.

Lesson 1 • Evaluation in Production Conditions

Simulates real-world evaluation using holdout sets, A/B tests, and shadow deployments. Production evaluation reveals failures that offline metrics miss.

Lesson 2 • Overfitting, Underfitting, and Generalisation

Diagnoses bias-variance tradeoff issues using learning curves and validation gaps. Generalisation ability determines whether a model succeeds in production.

Lesson 3 • Core Evaluation Metrics

Defines accuracy, precision, recall, F1, AUC, and regression metrics. Metric choice must align with business objectives and class imbalance conditions.

Lesson 4 • Fairness and Bias Evaluation

Applies fairness metrics to detect disparate model performance across groups. Fairness evaluation is required before deployment in high-stakes contexts.

Lesson 1 • Serving Architectures and APIs

Designs REST and gRPC inference endpoints for real-time and batch serving. Serving architecture determines latency, throughput, and integration complexity.

Lesson 2 • Scalability and Infrastructure

Configures auto-scaling, load balancing, and hardware acceleration for AI workloads. Infrastructure decisions directly affect cost and user experience at scale.

Lesson 3 • Deployment Strategies and Rollout

Applies blue-green, canary, and rolling deployment patterns to minimise risk. Controlled rollout strategies protect users from model regressions during releases.

Lesson 4 • CI/CD Pipelines for AI

Automates testing, validation, and deployment through continuous integration pipelines. CI/CD reduces manual errors and accelerates safe model releases.

Lesson 1 • Model Retraining Strategies

Designs scheduled and trigger-based retraining pipelines to keep models current. Retraining strategy balances freshness, cost, and deployment risk.

Lesson 2 • Data and Concept Drift Detection

Detects statistical shifts in input distributions and target relationships over time. Drift detection triggers retraining before user-facing performance degrades.

Lesson 3 • Model Governance and Lifecycle Management

Tracks model versions, ownership, and retirement decisions across the organisation. Governance ensures accountability and prevents unmanaged model proliferation.

Lesson 4 • Production Monitoring Fundamentals

Establishes metrics, dashboards, and alerting for live AI system health. Monitoring is the first line of defence against silent model failures in production.

Lesson 1 • Privacy and Data Protection

Implements privacy-preserving techniques and data minimisation in AI pipelines. Privacy controls protect individuals and reduce organisational liability.

Lesson 2 • AI Ethics Principles and Frameworks

Surveys established ethical principles including fairness, accountability, and transparency. Frameworks provide structured guidance for responsible design decisions.

Lesson 3 • Regulatory Compliance for AI

Maps AI system attributes to functional compliance requirements across sectors. Compliance readiness prevents deployment delays and regulatory penalties.

Lesson 4 • Risk Assessment for AI Systems

Applies risk identification and scoring methods to AI deployment scenarios. Risk assessment determines safeguards required before a system goes live.

Lesson 1 • Integration and End-to-End Testing

Validates the full AI system through integration, regression, and user acceptance tests. End-to-end testing catches failures that unit tests miss in complex pipelines.

Lesson 2 • Rapid Prototyping and Iteration

Builds minimum viable AI prototypes to validate assumptions before full development. Rapid iteration reduces wasted effort on unvalidated design choices.

Lesson 3 • Problem Framing and Requirements Definition

Translates business needs into precise AI problem statements and success criteria. Clear requirements prevent scope creep and misaligned model objectives.

Lesson 4 • System Architecture Design

Designs end-to-end AI system architecture including data, model, and serving layers. Architecture decisions made here propagate through all implementation phases.