MLOps

MLOps is the set of practices for keeping machine-learning models working reliably after they ship, the ML version of DevOps. A model isn’t like ordinary code that keeps doing the same thing forever: it learned patterns from past data, and when the world shifts, new customer behavior, a new product, a pandemic, its predictions quietly get worse even though nothing in the code changed. This is called drift. MLOps covers versioning data and models, automating retraining and deployment, and monitoring live inputs and accuracy so you notice drift before your users do. Much of real-world ML engineering is this work rather than inventing new models.