Machine Learning Automation through DevOps

Source: NESA Software Engineering Course Specifications (page 27).

MLOps is the automated process of designing, training and deploying machine learning models. It borrows many of the same principles and practices used in DevOps, bringing together the teams involved in developing machine learning models and the operational teams involved in deploying and supporting the models in production.

Students should know the three stages of MLOps.

Design

• defining the business problem to be solved
• refactoring the business problem into a machine learning problem
• defining success metrics
• researching available data.

Model development

• data wrangling
• feature engineering
• model training
• model testing and validation.

Operations

• model deployment
• supporting operations/use
• monitoring model performance.