Describe How to Access and Move Data for AI Workloads CERTIFIED-AI-EXPERT Practice Questions (Page 3)

Questions 11–15

1. 11application · medium

   A retail company collects sales data in CSV format from thousands of stores. The data contains dates in different formats and prices in different currencies. The AI pipeline must standardize these fields before training. During the data movement process, which step is essential?

   • AA step that renames files to include the store ID
   • BA step that splits the data into training and validation sets
   • CA preprocessing step that parses dates and converts currencies to a common standard
   • DA compression step to reduce the file size
   Reveal Solution Discussion Select an answer first
2. 12application · medium

   A media company stores video files in an S3-compatible object store and wants to process them with a GPU cluster that only supports POSIX file access. The processing job needs to read files sequentially and does not require random access. Which approach should they take to make the data accessible to the GPU cluster?

   • AMount the S3 bucket directly via the S3 protocol
   • BUse SMB to access the S3 bucket
   • CDownload the files to local disk using S3 API calls and then read them
   • DUse an S3-to-NFS gateway that presents the object store as a file system
   Reveal Solution Discussion Select an answer first
4. 13foundation · easy

   Which data transformation step is commonly applied during data movement to ensure that features from different sources have a consistent scale for machine learning models?

   • ANormalization
   • BEncryption
   • CCompression
   • DFormat conversion
   Reveal Solution Discussion Select an answer first
5. 14application · medium

   A media company generates large video files daily and needs to move them from a production site to a cloud data lake for model training. The transfer window is limited to off-peak hours, and the company wants to ensure only new files are transferred to avoid redundant data movement. Which strategy should they use?

   • AUse an incremental sync tool that transfers only changed or new files
   • BPerform a full copy of all files every day
   • CUse a manual copy process triggered by an operator
   • DStream the files in real-time as they are created
   Reveal Solution Discussion Select an answer first
6. 15expert · hard

   A company runs a distributed training job on a Kubernetes cluster. The training data is stored in an S3 bucket, and the training pods need to read the data. The data is read-heavy with many random accesses. The company wants to minimize the time to start training and avoid storing a full copy of the data in the cluster. Which approach should they use?

   • AUse a CSI driver that mounts the S3 bucket as a POSIX file system
   • BUse a sidecar container that downloads the data to the pod's local disk at startup
   • CPre-copy the entire dataset to a persistent volume before starting training
   • DExpose the S3 bucket via an NFS server running in the cluster
   Reveal Solution Discussion Select an answer first