MLD108 · Putting Models to Work: Deployment and Monitoring

Categories: AI & Automation
Wishlist Share

About Course

Turn a trained model into a dependable service or batch job, track experiments and versions, and monitor data drift and performance so the model keeps earning its place.

Advance · 8 hours · 4 lessons, a lab, an assessed assignment and a portfolio piece

Who it is for

Data practitioners and developers who need to put machine learning models into everyday business use and keep them healthy.

Prerequisites

MLD105 and MLD107; basic command-line and Git skills.

What you will be able to do

  • Choose between batch scoring and a real-time API and package a scikit-learn pipeline for either
  • Track experiments, data versions and models with MLflow so any prediction can be traced to its model
  • Serve a model with FastAPI in a Docker container with input validation and health checks
  • Monitor data drift, prediction drift and performance, and define retraining and rollback rules

Lessons

  • Lesson 1 · From notebook to production
  • Lesson 2 · Tracking, versioning and automation
  • Lesson 3 · Monitoring and maintenance
  • Lesson 4 · Lab and assignment

Lab: Deploy and monitor a demand-forecast model

You will take a demand-forecast model for a small chain of bakeries, log it with MLflow, serve it through a FastAPI endpoint in Docker and run a nightly batch scoring script. You will then simulate two months of new data, detect drift with Evidently and decide whether to retrain.

Free tools: Docker Desktop, FastAPI, MLflow (local), Evidently (open source) and GitHub Actions free tier, on your own computer

Assessed assignment

Prepare a production plan and working prototype for a model of your choice or the provided bakery forecast. Deliver the packaged model with MLflow tracking, either a Dockerized API or a scheduled batch job with a clear justification, a monitoring report that covers input drift, prediction drift and performance against actuals, and a one-page operations document that states owners, alert thresholds, retraining triggers, rollback steps and how predictions are logged.

Criterion Weight What excellent work looks like
Packaging and serving 25% The model runs from documented commands as an API or batch job, validates input and reports its version.
Tracking and reproducibility 25% Experiments, data versions and model versions are tracked so any prediction can be traced and reproduced.
Monitoring 30% Drift and performance checks are implemented, thresholds are justified and results are interpreted correctly.
Operations document 20% Owners, alerts, retraining triggers and rollback steps are specific and realistic for a small team.

Portfolio piece

A deployed, version-tracked model with a monitoring report and an operations document.

Self-paced, 12-month access. Questions are answered in writing through the student panel. A completion record is issued when you finish the course.

Next step

Continue with MLD109 · Final Project: A Predictive Model for a Business Decision · 8 hours · $56.00

Continue to the next course

Prefer the whole program? Program in Machine Learning & Applied Data Science · $378.00 (10% less than buying the courses separately)

Show More

What Will You Learn?

  • Choose between batch scoring and a real-time API and package a scikit-learn pipeline for either
  • Track experiments, data versions and models with MLflow so any prediction can be traced to its model
  • Serve a model with FastAPI in a Docker container with input validation and health checks
  • Monitor data drift, prediction drift and performance, and define retraining and rollback rules

Course Content

MLD108 · Putting Models to Work: Deployment and Monitoring (8 h)
Turn a trained model into a dependable service or batch job, track experiments and versions, and monitor data drift and performance so the model keeps earning its place.

  • Lesson 1 · From notebook to production
  • Lesson 2 · Tracking, versioning and automation
  • Lesson 3 · Monitoring and maintenance
  • Lesson 4 · Lab and assignment
  • MLD108 Assessment · Putting Models to Work: Deployment and Monitoring
Scroll to Top