AI Concepts CERTIFIED-TECHNOLOGY-SOLUTIONS-PROFESSIONAL Practice Questions (Page 2)

Questions 6–10

1. 6application · medium

   A research lab is training a large language model on a cluster of GPU servers. The training dataset is stored on a network file system, and the GPUs are frequently idle waiting for data to be loaded. The lab wants to reduce data-loading bottlenecks without changing the dataset size. Which storage solution would best address this issue?

   • AIncrease the network bandwidth between the file server and the GPUs
   • BUse object storage with HTTP-based access
   • CReplace the network file system with a parallel file system designed for high-throughput access
   • DMove the dataset to a single large NVMe drive attached to one GPU server
   Reveal Solution Discussion Select an answer first
2. 7expert · hard

   A financial firm is considering deploying an AI model for credit scoring. The model must have very low latency for real-time decisions, and the firm has strict data governance policies that require data to be stored on-premises. However, the firm's on-premises infrastructure has limited GPU capacity, and they expect variable demand. Which deployment model is most appropriate?

   • APublic cloud deployment with a dedicated region
   • BHybrid deployment with training on-premises and inference in the cloud
   • CEdge deployment on local servers at each branch
   • DOn-premises deployment with a GPU cluster sized for peak demand
   Reveal Solution Discussion Select an answer first
4. 8application · medium

   A genomics research institute is storing large sequencing files that are accessed by multiple analysis pipelines. The files are written once and read many times. The institute wants a cost-effective storage solution that can scale to petabytes and supports concurrent access. Which storage architecture is best suited?

   • AObject storage
   • BTape storage
   • CA single network-attached storage (NAS) device
   • DLocal NVMe drives on each analysis server
   Reveal Solution Discussion Select an answer first
5. 9foundation · easy

   Which storage characteristic is most important for AI training workloads that read large datasets repeatedly?

   • ALow cost per gigabyte
   • BHigh throughput and low latency
   • CSmall block size support
   • DHigh durability
   Reveal Solution Discussion Select an answer first
6. 10application · medium

   A financial services company is developing a fraud-detection model that must process sensitive customer data. They need low-latency inference for real-time transactions, but strict data-residency regulations require that all customer data remain within the country. The company has limited on-premises GPU capacity and wants to avoid large capital expenditures. Which deployment model best meets these requirements?

   • AHybrid deployment with training on-premises and inference in the public cloud
   • BEdge deployment on local point-of-sale devices
   • CPublic cloud deployment in a region within the country
   • DOn-premises deployment with dedicated GPU servers
   Reveal Solution Discussion Select an answer first