Task 1.3: Implement Data Validation and Processing Pipelines for FM Consumption. AIP-C01 Practice Questions (Page 3)

Questions 11–15

1. 11foundation · easy

   A developer is building a pipeline that ingests audio files, converts them to text, and then passes the text to a foundation model for summarization. Which AWS service should be used to transcribe the audio files into text?

   • AAmazon Rekognition
   • BAmazon Polly
   • CAmazon Comprehend
   • DAmazon Transcribe
   Reveal Solution Discussion Select an answer first
2. 12expert · hard

   A company is using Amazon Bedrock to build a question-answering system over a large corpus of internal documents. The documents are stored in S3 and are in various formats (PDF, DOCX, TXT). The team has built a RAG pipeline that extracts text, chunks it, and generates embeddings. The system's accuracy is lower than expected. The team suspects the issue is with the quality of the text extraction, particularly for PDFs that contain complex layouts (multi-column, tables). What is the MOST effective way to improve the text extraction quality for these complex PDFs?

   • AConvert the PDFs to images and use Amazon Rekognition to detect the text in the images.
   • BUse a simple PDF parsing library in a Lambda function to extract the text, as it is faster and cheaper than Textract.
   • CUse Amazon Textract with the 'Forms' and 'Tables' features enabled to extract text while preserving the structure of the document.
   • DUse Amazon Bedrock to analyze the raw PDF file and generate a text summary of its contents.
   Reveal Solution Discussion Select an answer first
4. 13application · medium

   A developer is building an application that uses Amazon Bedrock to generate product descriptions from a set of attributes. The attributes are stored in a JSON file. The developer wants to format the input for the FM. The FM is a text generation model. What is the best way to construct the prompt?

   • ASend the raw JSON object as the prompt, as the model can understand structured data formats.
   • BUse a special token to separate the JSON data from the instruction, such as '[INST] Generate a product description [/INST] {json}'.
   • CConvert the JSON attributes into a natural language sentence, such as 'Generate a product description for a blue, wireless mouse that is ergonomic.'
   • DEncode the JSON attributes as a base64 string and include it in the prompt.
   Reveal Solution Discussion Select an answer first
5. 14application · medium

   A real estate company is building an application that allows users to search for properties using natural language. The application uses Amazon Bedrock to generate embeddings for property descriptions and images. The pipeline needs to process new property listings as they are added. Each listing has a text description and several images. What is the most efficient way to generate embeddings for these new listings?

   • AUse a SageMaker Processing job that runs on a schedule to scan the S3 bucket for new listings and generate embeddings in batches.
   • BUse Amazon Kinesis Data Firehose to stream the listing data to a Lambda function that generates embeddings and stores them in S3.
   • CUse a Lambda function triggered by an S3 event to invoke the Bedrock embedding model for the text and images, and store the resulting vectors in Amazon OpenSearch Serverless.
   • DUse an AWS Glue ETL job to transform the listing data and call the Bedrock embedding model from within the job.
   Reveal Solution Discussion Select an answer first
6. 15foundation · easy

   A developer is writing code to invoke an Amazon Bedrock foundation model. The API request body must include the model ID, the input text, and inference parameters. What data format does the Amazon Bedrock InvokeModel API expect for the request body?

   • AXML
   • BYAML
   • CJSON
   • DCSV
   Reveal Solution Discussion Select an answer first