Operationalizing Machine Learning and Generative AI Solutions practice questions
MLOps and GenAIOps engineers on Azure
What the AI-300 exam covers
Current study guideSample AI-300 questions
Five of the 25 practice questions, each with its answer and explanation.
Design and implement MLOps infrastructure
How can components and environments be shared across several Azure Machine Learning workspaces?
- With registries
- They cannot be shared
- By copying VMs
- By emailing files
Show the answer
A. With registries
Registries make assets available across workspaces.
Design and implement MLOps infrastructure
How should workspaces be created consistently across environments?
- As infrastructure as code with Bicep or the Azure CLI
- By cloning a laptop
- From memory
- By clicking through the portal each time
Show the answer
A. As infrastructure as code with Bicep or the Azure CLI
Code-based deployment is repeatable and reviewable.
Design and implement MLOps infrastructure
How should GitHub Actions authenticate to Azure without storing long-lived secrets?
- No authentication
- Workload identity federation using OpenID Connect
- A personal access token in the README
- A password in the workflow file
Show the answer
B. Workload identity federation using OpenID Connect
Federated credentials let workflows obtain short-lived tokens.
Design and implement MLOps infrastructure
Which compute choice suits training jobs that come and go?
- A single always-on GPU VM
- A laptop
- A database server
- A compute cluster that scales down to zero when idle
Show the answer
D. A compute cluster that scales down to zero when idle
Autoscaling clusters provide capacity when jobs run and cost nothing when idle.
Implement ML model lifecycle and operations
Which tool tracks parameters, metrics and models for training runs in Azure Machine Learning?
- Excel
- Azure DNS
- MLflow
- Azure Front Door
Show the answer
C. MLflow
Azure Machine Learning uses MLflow for experiment tracking.