You are creating a new Azure Machine Learning workspace for an MLOps team. Which Azure resource is automatically provisioned as a dependency when you create a Machine Learning workspace?
Ai300 Mlops Infrastructure Category
Question · Easy
Spot the AI hallucination
Short
An MLOps engineer needs to register an external Azure Blob Storage container as a datastore in an Azure Machine Learning workspace. Which authentication method is recommended for production workloads following least-privilege principles?
Question · Medium
Introduction & Exam Overview
1.1 Current Exam Facts
Study guide
Which Azure Machine Learning compute type is most cost-effective for a long-running training job that requires GPUs and scales out across multiple nodes?
Tech Exam Prep Podcast
Listen on Spotify
Podcast
You want to deploy an Azure Machine Learning workspace and its dependent resources using infrastructure as code. According to the AI-300 study guide, which two technologies should you use?
Which GitHub Actions trigger should an MLOps engineer use to retrain a model whenever new training data is committed to a specific folder in the repository?
To restrict network access to an Azure Machine Learning workspace so that traffic from the public internet is denied, which configuration should you apply?
Question · Hard
1.2 Skills Measured, Format, and How This Guide Is Organized
Which Azure Machine Learning entity packages the Python dependencies, base image, and conda specification needed for a job to run?
An MLOps engineer wants to share a curated training dataset across multiple Azure Machine Learning workspaces in different regions. Which feature should they use?
When configuring identity and access management for an Azure Machine Learning workspace, which built-in role grants the ability to submit jobs and manage assets but not modify workspace-level settings?
You need to capture a training script's parameters, metrics, and output artifacts in a way that is portable across MLOps tools. Which open-source framework does Azure Machine Learning natively integrate with for experiment tracking?
Ai300 Model Lifecycle Category
1.3 Candidate Profile, Azure ML v2, Foundry, and DevOps Tooling
Which Azure Machine Learning capability automatically explores algorithms, preprocessing steps, and hyperparameters to find a high-performing model for a given dataset and task?
An MLOps engineer must tune learning rate and batch size for a training script. Which Azure Machine Learning feature should they use to run hyperparameter sweeps efficiently?
You need to train a large deep learning model across 8 NVIDIA A100 GPUs spread over 2 nodes. Which configuration should you set in your Azure Machine Learning command job?
Which Azure Machine Learning concept lets you compose multi-step training workflows (data prep, train, evaluate, register) that are reproducible, parameterized, and runnable from CI/CD?
Chapter 2: Azure Machine Learning Workspaces, Datastores, Compute, and IAM
2.1 Create and Manage Azure Machine Learning Workspaces
After training, you want to register a model in the workspace so it can be referenced by deployment pipelines. Which model format provides the richest metadata, including signature and conda environment, for downstream deployment?
Your team must evaluate a fraud-detection model for fairness across age groups before production. Which Azure feature provides disaggregated fairness, performance, and error analysis dashboards?
Which Azure Machine Learning endpoint type is optimized for low-latency real-time scoring with a single REST request per call?
You deploy a new version of a model to a managed online endpoint. To validate it on 10 percent of live traffic before promoting, which deployment strategy should you configure?
2.2 Create and Manage Datastores
A nightly job needs to score 50 million rows from a parquet dataset and write predictions back to storage. Which Azure Machine Learning deployment pattern is most appropriate?
You need to detect when the distribution of incoming inference data drifts away from the training data distribution. Which Azure Machine Learning capability should you configure?
When packaging a feature retrieval specification with a model artifact in Azure Machine Learning, what is the primary benefit?
Which Microsoft Foundry resource type is the top-level container for AI projects, model deployments, and connections in a GenAIOps deployment?
Ai300 Genaiops Infrastructure Category
2.3 Create and Manage Compute Targets
An MLOps engineer is automating Foundry project deployment with Bicep. Which identity type should they prefer for the project to authenticate to Azure OpenAI and Azure AI Search?
Which deployment option for an Azure OpenAI model offers a serverless pay-per-token API with no compute reservation?
A team requires guaranteed throughput and predictable latency for a high-volume Azure OpenAI workload. Which capacity option should they purchase?
When selecting a foundation model for a customer-support summarization use case, which two criteria should most influence the choice?
2.4 Configure Identity and Access Management for Workspaces
Which practice enables a team to track prompt versions, compare variants, and roll back to an earlier prompt definition?
Your generative app has two prompt variants: A (concise) and B (verbose). You want to compare quality on a fixed evaluation set. Which Foundry capability should you use?
Which network configuration prevents data exfiltration from a Foundry hub by ensuring traffic to dependent services traverses the Microsoft backbone instead of the public internet?
Which built-in evaluator in Microsoft Foundry measures whether a generated answer is supported by the retrieved context in a RAG application?
Ai300 Quality Observability Category
Chapter 3: Azure Machine Learning Data Assets, Environments, Components, and Registries
3.1 Create and Manage Data Assets
Which Microsoft Foundry evaluator detects whether a generated response contains hateful, sexual, violent, or self-harm content?
You want to calculate domain-specific quality (e.g., adherence to a banking compliance checklist) for generated answers. Which approach should you take in Foundry?
Which Foundry feature lets you continuously monitor a deployed generative AI application for groundedness and safety on a sampled stream of production traffic?
Which two metrics are most useful for tracking the operational performance of a deployed generative AI app?
3.2 Create and Manage Environments
Which Foundry capability enables tracing of a single user request through every step of a generative AI application (retrieval, prompt construction, LLM call, post-processing)?
To control runaway cost on an Azure OpenAI workload, which combination should you implement?
A RAG application returns answers with poor relevance. Which retrieval-side change should you try first?
Ai300 Optimization Category
Loading more…