MLOps & AI Deployment at Scale

Course outline and learning expectations

This is a tutor-led course. The outline shows what your tutor will cover; teaching materials and examinations are provided directly to enrolled students.

What you will learn

• Explain the MLOps lifecycle and diagnose common reasons models fail in production
• Create reproducible repositories that version code, datasets, model artefacts and configuration
• Track experiments, metrics, parameters and model lineage
• Build automated tests and CI/CD workflows for training and deployment pipelines
• Package and serve a machine-learning model with Docker and a Python API
• Monitor service health, prediction behaviour, data quality and data drift
• Define alert thresholds, retraining triggers and rollback procedures
• Design offline-first inference and sync-later workflows for intermittent connectivity
• Compare batch, real-time, local, edge and cloud deployment options against cost and operational constraints
• Deliver and document a production-grade AgriTech crop-disease pipeline with monitoring and offline fallback

Modules

1. 1

   MLOps foundations and production failure modes

   Understand the machine-learning lifecycle beyond training and identify reliability, data, dependency, scaling and governance risks

   MLOps lifecycleDevelopment versus productionFailure modesReproducibilityTechnical debt
2. 2

   Version control for data and models

   Create traceable and reproducible projects using Git, DVC and experiment tracking

   Git workflowDataset versioningModel artefact versioningExperiment trackingLineageReproducible runs
3. 3

   CI/CD basics for ML projects

   Automate code quality, tests, data checks, model validation and controlled releases

   Unit and integration testsData validationModel acceptance checksGitHub Actions or GitLab CIDeployment gatesSecrets and configuration
4. 4

   Containerisation with Docker

   Package a Python model service consistently for local and server deployment

   Docker images and containersDockerfileDependency managementFastAPI model endpointDocker Compose basicsHealth checks
5. 5

   Model monitoring and drift detection

   Observe service health and changing data or prediction patterns after deployment

   LoggingLatency and error metricsData qualityData driftPrediction driftAlert thresholdsRetraining and rollback triggers
6. 6

   Offline-first and sync-later architectures

   Design resilient AI services that continue operating during network interruptions and synchronise safely later

   Local inferenceQueued requests and resultsCachingRetry strategiesConflict handlingSecure synchronisationFallback behaviour
7. 7

   Cost-efficient deployment on limited budgets

   Select practical infrastructure and operating patterns that balance reliability, performance and cost

   Batch versus real-time inferenceCPU-friendly servingRight-sizingCachingOpen-source toolingStorage and network costsCost monitoring
8. 8

   Capstone — production-grade AgriTech ML pipeline

   Build, test, containerise, monitor and document a crop-disease pipeline with an offline fallback mode

   Repository and pipeline designData and model versioningAutomated testsContainerised APIMonitoring and drift reportOffline fallbackDeployment guideOperational runbook