Task Statement 3.1: Describe Design Considerations for Applications That Use Foundation Models (FMs). AIF-C01 Practice Questions (Page 2)

Questions 6–10

1. 6application · medium

   A startup is building a legal document summarization tool. They have a small dataset of labeled summaries and a limited budget. They need the model to follow a specific format for the summaries. Which customization approach is most cost-effective and appropriate?

   • AUse in-context learning with a few examples in the prompt to show the desired format.
   • BFine-tune the model on the small labeled dataset.
   • CUse RAG to retrieve legal documents and include them in the prompt.
   • DPre-train a custom foundation model from scratch on legal documents.
   Reveal Solution Discussion Select an answer first
2. 7application · medium

   A logistics company wants to automate the process of handling customer service tickets. The system should be able to read an incoming email, determine the issue type, and if it is a refund request, check the order database and initiate a refund. Which AWS AI service or feature would best enable this multi-step workflow?

   • AUse Amazon Textract to extract the email content and then use a rule-based system to process refunds.
   • BUse an AI agent that can call multiple tools, such as a database lookup and a refund API, to complete the task end-to-end.
   • CUse Amazon Comprehend to detect the sentiment of the email and automatically trigger a refund if sentiment is negative.
   • DUse a single foundation model with a very large context window to process the email and output the refund decision.
   Reveal Solution Discussion Select an answer first
4. 8application · medium

   A healthcare company wants to build an application that answers patient questions based on their internal medical guidelines. The guidelines are stored as PDFs and frequently updated. The company wants to avoid fine-tuning the model each time the guidelines change. Which approach should they use?

   • AUse Amazon Bedrock Knowledge Bases to ingest the PDFs, store embeddings in a vector database, and use RAG to retrieve relevant passages at query time.
   • BUse a pre-trained model that already contains medical knowledge and rely on its internal knowledge to answer questions.
   • CFine-tune the foundation model on the PDFs so the model memorizes the guidelines and can answer without external retrieval.
   • DStore the PDFs in Amazon S3 and use the foundation model to read them directly at query time.
   Reveal Solution Discussion Select an answer first
5. 9application · medium

   A developer is using a foundation model to generate code. They notice that the model sometimes produces code with syntax errors. They want to reduce the number of errors while still allowing some variation in the generated code. Which inference parameter adjustment is most appropriate?

   • AIncrease the temperature to a higher value, such as 0.8.
   • BDecrease the temperature to a lower value, such as 0.2.
   • CDecrease the max output length to force shorter code.
   • DIncrease the max output length to allow longer code snippets.
   Reveal Solution Discussion Select an answer first
6. 10application · medium

   A company wants to automate the process of summarizing meeting notes and sending action items to attendees. They plan to use an AI agent. What is the primary role of the agent in this scenario?

   • ATo generate a summary in a single model call without any external actions.
   • BTo orchestrate multiple steps, such as summarizing the notes and sending emails.
   • CTo replace the need for any human involvement in the process.
   • DTo store meeting notes in a database.
   Reveal Solution Discussion Select an answer first