Design and Implement Retrieval-Augmented Generation (RAG) DP-800 Practice Questions (Page 2)

Questions 6–10

1. 6expert · hard

   A human resources department stores employee performance reviews in a SQL database. They want to use an Azure OpenAI model to generate a summary of each review. The reviews contain sensitive personal data, and the company requires that the data be processed in a specific Azure region to comply with data residency regulations. What should you consider when designing the solution?

   • AUse a model that is hosted on-premises to avoid any cloud processing.
   • BUse a model that is trained on the employee data to avoid sending data to the cloud.
   • CEnsure the Azure OpenAI resource is deployed in the required region and that the database and the model are in the same region.
   • DEncrypt the data before sending it to the model to ensure compliance.
   Reveal Solution Discussion Select an answer first
2. 7foundation · easy

   A company stores product descriptions in a SQL database. They want to build a chatbot that answers customer questions about product features using the most current descriptions. Which approach best uses retrieval-augmented generation?

   • ARetrieve relevant product descriptions from the database and include them in the prompt to the language model
   • BSend the entire product catalog to the language model in every prompt without any retrieval
   • CTrain a custom language model on the product descriptions and then use that model for the chatbot
   • DUse a language model to generate product descriptions from scratch and store them in the database
   Reveal Solution Discussion Select an answer first
4. 8expert · hard

   You are building a RAG solution where the language model sometimes returns responses that are not in the expected JSON format, causing errors when you try to parse them. What is the best way to handle this?

   • AAdd instructions in the prompt to always return valid JSON and implement error handling to retry or log failures.
   • BUse a different language model that is guaranteed to return valid JSON.
   • CAssume the model will always return valid JSON and skip error handling.
   • DUse JSON_QUERY to force the response into valid JSON.
   Reveal Solution Discussion Select an answer first
5. 9foundation · easy

   You need to convert a result set from a SQL query into a JSON array to send to a language model. Which approach is appropriate?

   • AUse the CAST function with the AS JSON option
   • BUse JSON_QUERY to wrap the result set in a JSON array
   • CUse FOR JSON AUTO in the query to automatically format the result set as a JSON array
   • DUse OPENJSON with a specified schema to convert the result set to JSON
   Reveal Solution Discussion Select an answer first
6. 10expert · hard

   A financial services company wants to provide a natural-language query interface over its transaction database. The data is highly sensitive and subject to strict data-residency requirements. The company also needs to minimize the risk of data leakage to the language model provider. Which approach best balances these constraints?

   • AFine-tune a language model on the entire transaction database so it can answer without sending data at query time.
   • BUse a public language model endpoint without any data controls.
   • CExport the database to a CSV file and upload it to the language model provider for analysis.
   • DUse RAG with a language model endpoint hosted in the same Azure region and restrict the prompt to only the necessary retrieved records.
   Reveal Solution Discussion Select an answer first