Complete Notes

Course: DevOps, DataOps, MLOps — Duke University (Coursera)

Contents

Part Topic
1 What MLOps is
2 Foundations: DevOps, DataOps and how they relate
3 Core MLOps concepts
4 ML-specific components
5 Choosing the right approach
6 Software engineering practices (CI/CD, testing, tooling, IaC)
7 Containers and Kubernetes
8 Serving applications (frameworks, servers, microservices)
9 Cloud landscape
10 Hands-on examples from the module
11 Edge ML and model portability
12 Security, governance, sustainability
13 Monitoring and the production mindset
14 Summary (flow + table)

Part 1 — What MLOps Is

MLOps = applying DevOps, Software Engineering, and Data Operations principles to Machine Learning systems.

The goal is not just to train a good model. The goal is an ML system that is:

The ML System Lifecycle

Business Problem
      |
      v
    Data
      |
      v
Data Validation
      |
      v
Preprocessing / Feature Engineering
      |
      v
   Training
      |
      v
  Evaluation
      |
      v
Model Registry
      |
      v
 Deployment
      |
      v
   Serving
      |
      v
 Monitoring
      |
      v
Drift / New Data
      |
      v
 Retraining  ----> (loops back)

This is what "end-to-end MLOps" means: operationalization covers the entire lifecycle, not just deployment.

MLOps Is an Intersection