Task 1.5: Design Retrieval Mechanisms for FM Augmentation. AIP-C01 Practice Questions (Page 1)

Questions 1–5

1. 1foundation · easy

   A developer needs to generate embeddings for a large collection of documents stored in an Amazon S3 bucket. The developer wants to use AWS Lambda to process the documents in batches. What is the primary benefit of using Lambda for batch embedding generation?

   • ALambda provides automatic scaling to handle large volumes of documents without managing servers.
   • BLambda guarantees the lowest possible cost for embedding generation compared to other AWS services.
   • CLambda provides a built-in vector database for storing the generated embeddings.
   • DLambda eliminates the need to write code for embedding generation.
   Reveal Solution Discussion Select an answer first
2. 2application · medium

   A research organization is indexing scientific papers. Each paper has an abstract, sections, and references. They want to retrieve specific findings (e.g., 'What was the sample size?') and also support questions about the paper's overall methodology. They are using Amazon Bedrock Knowledge Bases. What chunking strategy should they use?

   • AChunk the entire paper as a single embedding.
   • BChunk by fixed token count (e.g., 500 tokens) with no overlap.
   • CChunk by paragraphs and use a sliding window with overlap.
   • DChunk by sections (e.g., abstract, methodology, results) and store each section as a separate chunk.
   Reveal Solution Discussion Select an answer first
4. 3foundation · easy

   A developer is designing a Lambda function to generate embeddings for documents in batches. The function reads a batch of documents from Amazon S3, calls an embedding model, and stores the resulting vectors. What is a key consideration when using Lambda for this batch processing?

   • ALambda automatically retries failed invocations, so no error handling is needed.
   • BLambda requires a VPC configuration to access Amazon S3.
   • CLambda functions cannot call Amazon Bedrock APIs.
   • DLambda has a maximum execution time limit, so the batch size must be tuned to fit within the timeout.
   Reveal Solution Discussion Select an answer first
5. 4application · medium

   A startup is building a RAG application on AWS. They have a relational database with structured product data and also unstructured PDF manuals. They want to enable semantic search over both types of data. They are already using Amazon Aurora PostgreSQL. What is the most efficient way to add vector search?

   • AUse Amazon Bedrock Knowledge Bases with Aurora as the vector store.
   • BEnable the pgvector extension on their existing Aurora PostgreSQL cluster and store embeddings in a new table.
   • CStore embeddings in Amazon DynamoDB and use a separate search service.
   • DUse Amazon OpenSearch Service as a separate vector store and synchronize data from Aurora.
   Reveal Solution Discussion Select an answer first
6. 5application · medium

   A healthcare startup is building a semantic search system over clinical trial documents. They need to compare different embedding models. They have a small team with limited ML expertise and need to minimize infrastructure overhead. They are already using Amazon Bedrock. What is the most appropriate approach for embedding model selection?

   • AUse a simple TF-IDF vectorizer from scikit-learn on a single EC2 instance.
   • BDeploy a custom Sentence-BERT model on Amazon SageMaker for maximum control.
   • CUse Amazon Titan Text Embeddings via Bedrock, selecting the appropriate model ID based on dimensionality and domain fit.
   • DUse Amazon Comprehend's built-in embeddings API.
   Reveal Solution Discussion Select an answer first