A university is building a RAG solution to answer questions from research papers. They notice that the retrieved chunks often contain irrelevant information, leading to poor answers. They want to improve retrieval precision without changing the embedding model. What should they do?

Increase the number of chunks retrieved and pass them all to the model.

Reduce the chunk size to make chunks more specific.

Use a hybrid search that combines keyword and vector search.

Remove the top-k parameter and use all chunks.

A university is building a RAG solution to answer student questions about course policies. They have the policies in PDF documents stored in Azure Blob Storage. They need to create an index for retrieval. Which approach should they use?

Load the PDFs directly into the prompt flow and use the LLM to search through them.

Convert the PDFs to text and store them in a single prompt template.

Use Azure SQL Database to store the PDFs and query them with SQL in the flow.

Use Azure AI Search to index the PDFs and connect the index to the prompt flow in Foundry.

A manufacturing company is planning to implement a generative AI solution for internal knowledge management. They have limited Azure experience and want to minimize operational overhead. They need to deploy a hub, project, and a model. Which planning approach should they take?

Create a hub and project using ARM templates, then deploy a model to an Azure Kubernetes Service cluster.

Use Azure CLI scripts to manually create all resources and deploy a model on a virtual machine.

Use the Microsoft Foundry portal to create a hub and project, and select a managed model deployment.

Use the Foundry SDK to create the hub and project, and then deploy a model to a local Docker container.

In a prompt template, what is the purpose of a placeholder like {{user_input}}?

It indicates a comment in the template

It defines the model's temperature setting

It is replaced with a dynamic value at runtime

A team has built a prompt flow that generates code snippets. They want to evaluate the flow's correctness and safety before deployment. They have a dataset of programming problems with known solutions. What is the best evaluation strategy?

Manually review the generated code for a few samples.

Deploy the flow and monitor for errors in production.

Use a general text quality metric like BLEU to evaluate the code.

Run the flow on the dataset and compare the generated code against the known solutions using automated tests.

Microsoft Certified:Azure AI Engineer Associate

Domain 2Objective 1

Build Generative AI Solutions with Microsoft Foundry AI-102 Practice Questions (Page 7)

Part of the Implement generative AI solutions domain, which accounts for 15-20% of the AI-102 exam. Microsoft does not publish an official question count, but from its 100-minute exam (~40–65 total, ~6–13 in this domain), expect 2–4 from this objective — we provide 45 practice questions to prepare you well beyond it. (estimate)

45questions here

9free pages

8concepts

15-20%of the exam

Questions 31–35

31application · medium
A university is building a RAG solution to answer questions from research papers. They notice that the retrieved chunks often contain irrelevant information, leading to poor answers. They want to improve retrieval precision without changing the embedding model. What should they do?
Select an answer first
32application · medium
A university is building a RAG solution to answer student questions about course policies. They have the policies in PDF documents stored in Azure Blob Storage. They need to create an index for retrieval. Which approach should they use?
Select an answer first
33application · medium
A manufacturing company is planning to implement a generative AI solution for internal knowledge management. They have limited Azure experience and want to minimize operational overhead. They need to deploy a hub, project, and a model. Which planning approach should they take?
Select an answer first
34foundation · easy
In a prompt template, what is the purpose of a placeholder like {{user_input}}?
Select an answer first
35expert · hard
A team has built a prompt flow that generates code snippets. They want to evaluate the flow's correctness and safety before deployment. They have a dataset of programming problems with known solutions. What is the best evaluation strategy?
Select an answer first

Finished these 5 questions?

Review the revealed explanations, or continue through the curriculum.

Next 5 questions →Back to AI-102

Free Basic Practice is a study aid with revealable answers — not a scored exam. Examers.io is independent and not affiliated with or endorsed by Microsoft. “AI-102” is a trademark of its owner, used for identification only.