Managed ML Platform across clouds
The managed notebook and training platform - SageMaker, Azure Machine Learning, Vertex AI, OCI Data Science - with the notebook off the internet and the data it reads encrypted with your key.
4 verified modules, 1 of them live-tested apply→verify→destroy; the rest are static-validated, live-test pending.
Compare by provider
| Provider | Module | Verification |
|---|---|---|
| AWS | A Notebook that Is Not Root with Open Egress | static-validated |
| Azure | An ML Workspace that Is Private, Isolated and Encrypted with Your Key | static-validated |
| Google Cloud | Vertex AI Endpoint | ✓ live-tested |
| Oracle Cloud | Notebook Sessions on Your Subnet that Bill Only When Meant To | static-validated |
How to choose
Compare the network posture first: whether the notebook gets a public address by default, whether it reaches the internet, and whether training jobs run in your VPC. Then what identity the notebook runs as and how wide it is, because a notebook is a shell with the platform credentials of whoever created it.
When not to use
The expensive part of these services is the compute left running, not the platform fee. A notebook instance that idles overnight costs the same as one that trains; every module here needs an idle-shutdown or a schedule from the buyer.