Azure · Cloud service
Azure Machine Learning with OrchestrAI
Catalog exported 2026-09-02
Set up Azure Machine Learning from chat: workspaces, compute, training, sweep and AutoML jobs, endpoints.
OrchestrAI exposes 7 Azure ML operations: 7 create or modify resources and run only after you confirm the plan.
What teams use it for
Data science teams use OrchestrAI to go from nothing to a running experiment: a workspace, a compute instance, and a command job submitted with the training script and environment named in plain language. Sweep jobs for hyperparameter search and AutoML jobs for a quick baseline follow the same pattern, and a managed online endpoint can be created when a model is ready to serve. Every operation here creates something, with no list, cancel, or delete and no step to attach a model to an endpoint, so monitoring jobs and finishing a deployment happen in Azure ML studio.
Every Azure ML operation, with its risk level
| Operation | What it does | Risk | Step-level approval |
|---|---|---|---|
Create Azure ML AutoML Job |
Submit an Azure ML AutoML job | Creates resources | No |
Create Azure ML Compute Instance |
Create an Azure ML compute instance | Creates resources | No |
Create Azure ML Online Endpoint |
Create an Azure ML managed online (scoring) endpoint | Creates resources | No |
Create Azure ML Registry |
Create an Azure ML registry (model registry) | Creates resources | No |
Create Azure ML Sweep Job |
Submit an Azure ML sweep (hyperparameter tuning) job | Creates resources | No |
Create Azure ML Training Job |
Submit an Azure Machine Learning command (training) job | Creates resources | No |
Create Azure ML Workspace |
Create an Azure Machine Learning workspace | Creates resources | No |
Risk tiers come from the catalog: low is read-only or low-impact, medium creates resources and is reversible, high modifies existing resources, destructive may lose data. Every plan that creates or changes resources is shown with its cost estimate and waits for your confirmation. Operations marked with a step-level approval pause again on their own step. Destructive operations require a typed risk phrase.
Prompts that work
- Create an Azure ML workspace called ml-acme in rg-ml in westus2
- Submit a training job in ml-acme that runs train.py on the cpu-cluster compute with the sklearn environment
- Start a sweep job over learning rate 0.001 to 0.1 and batch size 32 or 64 using the same script
Before anything runs
Every mutation shows its plan, cost estimate, and blast radius, then waits for your confirmation. Destructive operations require a typed risk phrase. Credentials are minted per run through OIDC federation and discarded afterward; nothing you create here is invisible later, because every resource lands in the desired-state ledger where drift is detected and can be converged. Details on the security page.
Frequently asked questions
- Does OrchestrAI need confirmation to submit an Azure ML training job?
- Training, sweep, and AutoML jobs are rated medium because they start billable compute but create no lasting resources, and OrchestrAI shows the job plan before submitting it.
- Can OrchestrAI deploy a model to an Azure ML endpoint?
- It can create the managed online endpoint, but attaching a model deployment to that endpoint is not an available operation yet.
- Which Azure ML operations need an extra approval step?
- None of the Azure ML operations currently carries a step-level approval gate; they are read-only or run after plan confirmation like any other change.
Other Azure services
Related integrations
Try it on your own account
Connect your cloud read-only and see your resources, drift, and costs before anything runs. $5 minimum to start. Unused credits refunded in your first 14 days.
Unused credits refunded in your first 14 days.