Start here in MLOps
The most useful guides in this topic, not just the most recent.
MLOps Guides & Tutorials
Operationalise machine learning — MLflow, Docker, Kubernetes, CI/CD, monitoring and production ML system design.
MLOps articles
MLOps vs DevOps vs DataOps: Key Differences
DevOps, DataOps and MLOps share a philosophy, automation and fast feedback, but they manage different things that fail in different ways. This guide compares them side by side, explains what machine learning adds that DevOps alone does not handle, and shows how the three fit together in a real team.
Model Registry: Managing the ML Model Lifecycle
A model registry is where trained models become managed releases. This guide runs a full lifecycle in MLflow: a production champion, a retrained challenger, a validation gate on held-out data, promotion by moving an alias, tags for audit, and a one-line rollback, with the reasoning behind each step.
Model Versioning with MLflow and DVC
A model version is only useful if you can say what produced it. This guide combines DVC for data with MLflow for models, so each registered version records its data hash and Git commit, then rebuilds an old version from that record alone and checks that the predictions come out identical.
Testing Machine Learning Code and Data
ML code can pass every unit test and still ship a broken model, because data and model behaviour need tests of their own. This guide writes nine pytest checks for a real model, covering schema, invariants, plausibility, a baseline gate and behaviour, and one catches a genuine sensor fault in the data.
A/B Testing ML Models in Production
A model that wins offline still has to prove itself with real customers. On real bank marketing outcomes, this guide shows why a one-point offline gain needed about 36,000 calls per arm to confirm, how to split traffic and check the split, and how checking results early turned 4% false alarms into 18%.
Blue-Green Deployments for ML Services
Blue-green deployment keeps the old model serving while the new one waits beside it, so switching either way is quick. This guide runs it with two real spam filters, shows a pre-switch check catching a score-scale change that would have made a review queue five times longer, and switches under live traffic.