AI Azure AI Engineer 4 — Questions and Answers
Question 1: Which Azure OpenAI parameter controls how deterministic or random the model's output is during text generation?
- max_tokens
- top_p
- temperature (Correct answer)
- frequency_penalty
Correct answer: temperature
Temperature scales the probability distribution over tokens; lower values make output more deterministic, higher values increase randomness.
Question 2: A company needs real-time speech transcription with speaker diarization in a call center application. Which Azure service should they use?
- Azure Video Indexer
- Azure AI Speech with conversation transcription (Correct answer)
- Azure AI Language
- Azure Translator
Correct answer: Azure AI Speech with conversation transcription
Azure AI Speech's conversation transcription feature provides real-time multi-speaker diarization that identifies and labels each speaker's segments.
Question 3: What is the key advantage of using Azure OpenAI's function calling feature in an agent application?
- It increases the model's context window
- It enables the model to output structured JSON to invoke defined functions reliably (Correct answer)
- It reduces token consumption automatically
- It provides built-in memory across sessions
Correct answer: It enables the model to output structured JSON to invoke defined functions reliably
Function calling lets you define schemas for external functions and the model returns structured JSON arguments, enabling reliable tool use in agents.
Question 4: In Azure AI Search, what does the 'semantic ranker' do differently from standard BM25 ranking?
- It indexes documents faster using GPU acceleration
- It re-ranks BM25 results using language model-based semantic understanding (Correct answer)
- It filters results by geographic proximity
- It converts queries to SQL before searching
Correct answer: It re-ranks BM25 results using language model-based semantic understanding
Semantic ranker applies a cross-encoder model to re-score top BM25 candidates based on deeper semantic relevance to the query.
Question 5: Which Azure Machine Learning feature automatically explores multiple algorithms and hyperparameters to find the best model for a given dataset?
- Azure ML Pipelines
- Automated ML (AutoML) (Correct answer)
- Azure ML Designer
- Responsible AI Dashboard
Correct answer: Automated ML (AutoML)
AutoML iterates through algorithm and hyperparameter combinations, evaluating each against a specified primary metric to surface the best candidate model.
Question 6: When deploying a model to an Azure Machine Learning online endpoint, what does a 'blue-green deployment' strategy accomplish?
- It encrypts traffic using TLS 1.3 and TLS 1.2 simultaneously
- It routes a configurable percentage of traffic to a new model version while keeping the old version live (Correct answer)
- It deploys the model to two Azure regions for geo-redundancy
- It separates batch and real-time inference workloads
Correct answer: It routes a configurable percentage of traffic to a new model version while keeping the old version live
Blue-green (or canary) deployments on Azure ML online endpoints split live traffic between model versions, enabling safe gradual rollouts.
Question 7: Which Azure service should an AI engineer use to orchestrate a multi-step workflow that calls Azure AI services, stores results, and sends notifications—without writing server code?
- Azure Kubernetes Service
- Azure Logic Apps (Correct answer)
- Azure Container Instances
- Azure Service Bus
Correct answer: Azure Logic Apps
Azure Logic Apps provides a low-code workflow engine with connectors to hundreds of services including Azure AI, storage, and communication platforms.
AI Azure AI Engineer 4
Which Azure OpenAI parameter controls how deterministic or random the model's output is during text generation?