Build Generative AI Solutions with Microsoft Foundry AI-102 Practice Questions (Page 3)

Questions 11–15

1. 11foundation · easy

   Which resource is created within a Microsoft Foundry project to store and manage datasets?

   • AAzure Container Registry
   • BAzure Cognitive Search
   • CAzure Blob Storage
   • DAzure Key Vault
   Reveal Solution Discussion Select an answer first
2. 12application · medium

   A legal firm wants to build a generative AI solution that summarizes long court transcripts. The solution must support a context window large enough to process an entire transcript in one request, and the firm needs to compare the quality of two candidate models before committing. What should they do first?

   • AUse the model with the highest benchmark score on general summarization tasks.
   • BDeploy the model with the largest context window and start using it immediately.
   • CChoose the cheapest model that can handle the context window.
   • DEvaluate both candidate models using a representative set of transcripts and quality metrics.
   Reveal Solution Discussion Select an answer first
4. 13foundation · easy

   What is the primary consideration when selecting a generative AI model from the Microsoft Foundry model catalog?

   • AThe model's performance on the specific task
   • BThe model's name recognition
   • CThe model's training data size
   • DThe model's licensing terms
   Reveal Solution Discussion Select an answer first
5. 14expert · hard

   A company has a prompt flow deployed in Microsoft Foundry that is used by a web application. The application is experiencing high latency during peak hours. They want to reduce latency without changing the flow logic. What should they do?

   • AReduce the number of requests by adding a queue.
   • BUse a larger model to improve response quality.
   • CMove the flow to a different region.
   • DScale out the deployment by increasing the number of instances.
   Reveal Solution Discussion Select an answer first
6. 15expert · hard

   A legal tech company has a RAG solution that retrieves clauses from contracts. The evaluation shows high retrieval recall but low precision, and the final answers sometimes cite irrelevant clauses. They need to improve the quality of the final answers without changing the embedding model. What should they do?

   • ARemove the retrieval step and use the full contract as context.
   • BAdd a reranking step to the retrieval pipeline to select the most relevant chunks.
   • CIncrease the top-k value to retrieve more chunks.
   • DChange the prompt to instruct the model to ignore irrelevant clauses.
   Reveal Solution Discussion Select an answer first