ML Model Deployment Basics

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 how deployment moves a trained model from a notebook into a usable application
• Save and load compatible machine learning models using pickle or joblib and apply safe loading practices
• Build a prediction API with Flask or FastAPI
• Validate user inputs and return clear prediction responses
• Connect a model API to a basic mobile-friendly web form
• Prepare dependencies and configuration files for cloud deployment
• Deploy and test an ML web application on Render, Railway or PythonAnywhere where supported
• Optimise a simple interface for low-bandwidth users
• Record basic monitoring information such as errors, response time and prediction checks
• Deploy a working capstone model as a usable web tool

Modules

1. 1

   From Notebook to Real Application

   Understand deployment and prepare a trained model for use outside a notebook

   What model deployment meansTraining versus inferenceDeployment workflowProject structureChoosing a simple use case
2. 2

   Saving and Loading Models

   Persist and restore a trained model while managing compatibility and security risks

   pickle fundamentalsjoblib fundamentalsSaving preprocessing pipelinesEnvironment and version compatibilityNever loading untrusted model files
3. 3

   Building a Prediction API

   Create a simple HTTP prediction service with Flask or FastAPI

   API basicsRoutes and endpointsRequest validationLoading the model oncePrediction responsesLocal testing
4. 4

   Connecting the Model to a Web Page

   Build a basic interface that sends user inputs to the model and displays results

   HTML formsInput fieldsSubmitting requestsDisplaying predictionsError messagesMobile-friendly layout
5. 5

   Deploying on Accessible Platforms

   Prepare and publish the application using Render, Railway or PythonAnywhere, subject to current platform support and limits

   requirements.txtEnvironment variablesStart commandsGit and GitHub workflowRender deploymentRailway deploymentPythonAnywhere deploymentReading deployment logs
6. 6

   Designing for Low-Bandwidth Users

   Reduce page weight and make the tool practical for users with limited data or slower connections

   Lightweight HTML and CSSMinimal assetsCompressed images when neededClear formsFast responsesMobile browser testingOffline-friendly guidance
7. 7

   Basic Monitoring and Maintenance

   Check whether the deployed service and model continue to work as expected

   Health checksApplication logsError trackingResponse-time checksPrediction sanity checksInput drift awarenessUpdating dependencies safely
8. 8

   Capstone: Crop Disease Risk Web Tool

   Adapt the Course 10 crop-disease-risk model into a simple mobile-friendly web form that farmers can use in a phone browser

   Select and prepare the modelBuild the APICreate the web formDeploy the applicationTest on a phone browserDocument limitationsDemonstrate the final tool