Why MLOps
Why MLOps is needed, and the features, setup and core concepts of MLOps Central Platform.
As models, data and field devices multiply, it becomes hard to trace who built the model running in the field, when, and from what data. MLOps systematises building models repeatedly, deploying them safely and monitoring them in production. It is the model counterpart of DevOps for software. MLOps Central Platform provides this process as a single platform.
In this chapter
| Page | Time | What it covers |
|---|---|---|
| Problems in production | 3 min | Five reasons running models is hard, and how the platform responds |
| How it differs from MLflow | 4 min | Where MLflow alone falls short, and the layer the platform adds |
| What MLOps Central Platform does | 3 min | Features by menu |
| The setup at a glance | 3 min | Overall setup, users, and how the field connects to the centre |
| Deployment examples by situation | 5 min | Seven deployments, from a single factory to an air-gapped site |
| Before and after | 3 min | The same job done by hand and with the platform |
| Core concepts | 5 min | Tenants, datasets, experiments, model versions, approval, serving, edge, drift |
| Roles and permissions | 3 min | Permissions by role |