AI-300 Practice Questions 1 — 50 Questions

Question 1

You need to debug a training job that is failing inside a specific Docker container. What is the best way to investigate?

Accepted Answer

Use 'az ml job stream' to view logs. The 'az ml job stream' command streams the logs of a running or completed Azure ML job in real time, including container logs from the Docker environment. This allows you to see error messages, stack traces, and output from the training script inside the container, making it the most direct way to debug a failing job.

Answer

Delete the datastore

Answer

Use the Azure ML SDK to download the container image

Answer

Submit a new training job without parameters

Question 2

What is the primary function of an Azure ML 'Datastore'?

Accepted Answer

Connection to storage services. Datastores are abstractions to securely connect to underlying data storage services like Blob storage or Data Lake.

Answer

Experiment tracking

Answer

Model registry

Answer

Compute resource management

Question 3

Which THREE of the following are key components of a robust MLOps strategy in Azure?

Accepted Answer

Automated CI/CD pipelines. Automated CI/CD pipelines (B) are a core MLOps component because Azure DevOps or GitHub Actions can automatically build, test, and deploy ML assets via Azure Machine Learning pipelines, ensuring repeatable and reliable releases. Version control for code and models (C) is essential since Azure ML and Git track code, datasets, and model artifacts, enabling reproducibility, auditability, and rollback of experiments and deployments. Automated model monitoring (E) is also key because Azure Machine Learning model monitoring detects data drift, prediction drift, and performance degradation in production, triggering alerts or retraining. Manual model testing (A) is not a robust MLOps practice because it is error-prone, slow, and non-repeatable compared with automated testing in CI/CD. Using only local workstations (D) is incorrect because MLOps relies on shared, scalable cloud resources such as Azure Machine Learning compute, registries, and pipelines rather than isolated local environments.

Answer

Manual model testing

Answer

Using only local workstations

Question 4

A team is scaling a GenAI application that utilizes multiple models via model-as-a-service. You notice that inference requests are being throttled. What is the most effective infrastructure-level adjustment to handle the increased load?

Accepted Answer

Purchase and deploy Provisioned Throughput Units (PTUs) for the specific model deployment.. Provisioned throughput units (PTUs) allow for dedicated capacity, which is the standard way to prevent throttling for high-scale GenAI workloads in Azure.

Answer

Change the model version to a lower-parameter variant.

Answer

Move the deployment to a different subscription within the same Azure region.

Answer

Increase the number of instances in the associated Azure Kubernetes Service (AKS) cluster.

Question 5

You are deploying a model that requires a high-memory compute for inference. You are using a 'Managed Online Endpoint'. Where do you specify the instance type for this deployment?

Accepted Answer

In the 'Deployment' YAML file under 'instance_type'.. The instance type is defined in the 'Deployment' configuration object, typically in the 'instance_type' field.

Answer

In the 'Endpoint' YAML file.

Answer

In the 'Environment' definition.

Answer

In the 'Workspace' settings.

Question 6

You are deploying a GenAI app to production and want to track the 'Token Usage' and 'Latency' metrics per user session. Which service should you integrate with your application?

Accepted Answer

Azure Application Insights. Azure Application Insights is the APM service that captures custom metrics, request telemetry, dependencies, and traces, making it the right place to track per-session Token Usage and Latency. You can emit custom metrics (e.g., 'tokens_used', 'latency_ms') tagged with a session ID and visualize them in workbooks or query them with KQL.

Answer

Azure Container Registry

Answer

Azure Key Vault

Answer

Azure Policy

Answer

Azure AI Search

Question 7

Which component in an Azure ML pipeline definition is responsible for specifying the runtime environment, including Python packages and Conda dependencies?

Accepted Answer

Environment. In an Azure ML pipeline, the Environment object encapsulates the runtime (Docker base image), Python packages, and Conda dependencies used by a component. Components reference an Environment to define their execution context, so the Environment is the correct answer for specifying packages and Conda dependencies.

Answer

Compute Target

Answer

Datastore

Answer

Component

Question 8

You are configuring a batch scoring job. You need to ensure that the job processes data in parallel to reduce completion time. What property should you adjust in the 'ParallelRunConfig'?

Accepted Answer

Adjust 'process_count_per_node' and 'node_count'.. The 'node_count' and 'process_count_per_node' are the key parameters in 'ParallelRunConfig' that dictate the degree of parallelism.

Answer

Change the 'output_action' to 'append_row'.

Answer

Enable 'distributed_training'.

Answer

Set 'min_nodes' to 1.

Question 9

Which TWO of the following are valid ways to authenticate to an Azure AI Foundry project?

Accepted Answer

API Keys.. Option A (API Keys) is correct because Azure AI Foundry projects expose key-based authentication, where a project or resource key is passed in the request header (e.g., the api-key header) to authorize calls to the project's endpoints. Option B (Microsoft Entra ID with RBAC) is correct because Foundry projects support Entra ID token-based authentication, using role-based access control assignments (such as Cognitive Services User or Contributor) to authorize identities via OAuth 2.0 bearer tokens. Option C is not valid because hardcoding usernames in code is not an authentication mechanism and would be a security anti-pattern, not a supported credential type. Option D is not valid because local file-based passwords are not an authentication method offered by Azure AI Foundry. Option E is not valid because IP allow-listing is a network-level control, not an authentication method, and it does not verify caller identity.

Answer

Hardcoded usernames in the code.

Answer

Local file-based passwords.

Answer

Public IP allow-listing only.

Question 10

Which Azure AI Foundry feature allows you to evaluate your model's performance?

Accepted Answer

Evaluation. The Evaluation feature in Azure AI Foundry is specifically designed to assess model performance using metrics like accuracy, F1 score, and coherence. It provides built-in evaluators and supports custom evaluation flows, enabling systematic testing of models before and after deployment. This directly addresses the need to evaluate model performance, unlike other options which focus on monitoring, cataloging, or logging.

Answer

Endpoint monitoring

Answer

Model Catalog

Answer

Deployment logs

Question 11

You need to audit all model access logs across your organization. What is the most effective approach?

Accepted Answer

Enable 'Diagnostic Settings' to send logs to a 'Log Analytics Workspace'.. Enabling diagnostic settings on the Azure AI Services resource to send logs to a Log Analytics workspace is the standard for organization-wide auditing.

Answer

Review individual request headers in the app logs.

Answer

Use 'Azure Advisor' to detect unauthorized access.

Answer

Download the 'Model Usage' report from the billing console.

Question 12

A pipeline step fails because it cannot find a file in the datastore. What is the most likely cause?

Accepted Answer

Incorrect path definition in the component. The mount path or the relative path defined in the pipeline step component configuration is likely incorrect.

Answer

The workspace is too large

Answer

The compute cluster is off

Answer

The subscription is expired

Question 13

You notice that your fine-tuned model is 'forgetting' base capabilities after training on a small dataset. What strategy should you use to mitigate this?

Accepted Answer

Include a portion of the original training data during fine-tuning.. Catastrophic forgetting occurs when a model fine-tuned on a small dataset loses its general capabilities. To mitigate this, you should include a portion of the original training data (or a diverse set of general data) during fine-tuning so the model retains its base knowledge while learning the new task. This is a common technique in continual learning.

Answer

Increase the fine-tuning learning rate.

Answer

Change the model architecture to GPT-4.

Answer

Reduce the batch size to 1.

Question 14

Which parameter in the Azure OpenAI API should be adjusted to make the model's output more deterministic and repeatable?

Accepted Answer

temperature. Temperature controls the randomness of the model's output by scaling the logits before the softmax function. A lower temperature (e.g., 0 or near 0) makes the probability distribution sharper, so the model almost always picks the highest-probability token, resulting in deterministic and repeatable responses. Setting temperature to 0 is the standard way to achieve consistent outputs for the same input in Azure OpenAI.

Answer

max_tokens

Answer

presence_penalty

Answer

frequency_penalty

Question 15

You are deploying a model as a real-time endpoint. You want to ensure the deployment can handle sudden spikes in traffic. Which scaling configuration should you enable?

Accepted Answer

Auto-scaling. Auto-scaling allows the endpoint to dynamically adjust node count based on CPU/Memory usage.

Answer

Static node count

Answer

Batch deployment

Answer

Pre-warmed instances

Answer

None of the above

Question 16

You need to implement a retraining trigger based on performance degradation. Which TWO metrics should you monitor to decide when to retrain?

Accepted Answer

Model prediction precision. Options D and E are correct because retraining triggers based on performance degradation must directly measure how well the model's predictions match ground truth: model prediction precision (D) captures the proportion of positive predictions that are actually correct, and model prediction accuracy (E) captures the overall fraction of correct predictions, both of which decline when the model drifts or degrades. Monitoring these model-quality metrics lets you set thresholds (e.g., accuracy drops below 90% or precision falls by 5%) that signal when retraining is warranted. Option A (CPU usage of the inference cluster) is an infrastructure resource metric that reflects load or scaling needs, not model correctness. Option B (network latency) measures communication delay and says nothing about prediction quality. Option C (number of active users) is a traffic/usage metric that may correlate with load but does not indicate whether the model's predictions have degraded.

Answer

CPU usage of the inference cluster

Answer

Network latency

Answer

Number of active users

Question 17

You want to track your model experiments and compare their performance metrics. Which Azure ML feature provides this capability?

Accepted Answer

Experiments. In Azure Machine Learning, an experiment is the logical container for tracking runs of a model training script. Each run automatically logs metrics, parameters, artifacts, and outputs, enabling you to compare performance across different configurations. This is the core capability for experiment tracking and comparison.

Answer

Environments

Answer

Model Registry

Answer

Compute targets

Question 18

You notice your model is outputting redundant information. Which parameter specifically targets the penalty for repeating tokens?

Accepted Answer

frequency_penalty. The frequency_penalty parameter applies a penalty proportional to how many times a token has already appeared in the generated text, directly discouraging repeated tokens and reducing redundancy. Its value ranges from -2.0 to 2.0, with positive values increasing the penalty for frequent tokens.

Answer

max_tokens

Answer

presence_penalty

Answer

top_p

Question 19

You are configuring a connection to a vector store. What is the best way to handle the secret credential?

Accepted Answer

Store in Azure Key Vault and reference in the connection.. Key Vault is the centralized store for all application secrets in Azure, which should be linked to the project connection.

Answer

Hardcode the key in the flow configuration.

Answer

Pass it as a parameter in the code.

Answer

Store it in an environment variable.

Question 20

You are building a RAG application and notice that the model sometimes hallucinates information not present in the retrieved documents. Which evaluation metric should you prioritize to mitigate this?

Accepted Answer

Groundedness. Groundedness specifically assesses whether the generated response is derived from the retrieved documents.

Answer

Relevance

Answer

Performance

Answer

Fluency