Azure AI Engineer Flashcards
7 cards from real AI practice questions. Tap to flip, then mark Knew It or Still Learning — missed cards come back until you master them.
Read the first 7 Azure AI Engineer flashcards as text
A model trained on historical loan data consistently under-predicts default risk for a specific demographic group. This is an example of:
Answer: Algorithmic bias
Algorithmic bias occurs when a model produces systematically skewed outcomes for a protected group, often due to imbalanced or unrepresentative training data.
Which Azure tool helps detect when a deployed model's input data distribution shifts significantly from its training data?
Answer: Azure Machine Learning Data Drift Monitor
Azure ML Data Drift Monitor compares baseline training data statistics against live inference data and alerts when drift exceeds a configured threshold.
In the Azure AI Foundry prompt flow, what is the primary role of a 'tool node'?
Answer: Executing code or calling external APIs as a step in the flow
Tool nodes in prompt flow encapsulate executable logic—Python functions or API calls—that transform data between LLM steps.
Which Azure service provides pre-built AI models for document understanding, including invoice and receipt extraction, without custom training?
Answer: Azure Form Recognizer (Document Intelligence)
Azure AI Document Intelligence includes pre-built models for common document types like invoices, receipts, and ID documents that work out of the box.
When fine-tuning an Azure OpenAI model, what file format is required for the training dataset?
Answer: JSONL with chat completion format
Azure OpenAI fine-tuning requires JSONL files where each line is a JSON object with a 'messages' array following the chat completion schema.
An engineer wants to evaluate multiple LLM responses for groundedness against source documents at scale. Which Azure capability supports this?
Answer: Azure AI Evaluation SDK with GPT-based evaluators
The Azure AI Evaluation SDK provides built-in GPT-powered evaluators including groundedness, relevance, and coherence that can run at scale in pipelines.
Which feature of Azure Machine Learning allows you to reuse and share preprocessing steps and feature engineering logic across multiple experiments?
Answer: Pipelines with registered components
Registered components in Azure ML pipelines encapsulate reusable steps that can be versioned and shared across teams and experiments.