AI Platforms, Tools, and Ecosystem Integration CAIPM Practice Questions (Page 9)

Questions 41–45

1. 41application · medium

   A university research lab needs to run custom deep learning experiments that require specific GPU configurations and frequent code changes. They have limited budget and prefer to avoid long-term commitments. Which type of AI platform is most suitable?

   • AA fully managed enterprise AI platform with a one-year contract
   • BAn on-premises GPU cluster that requires a large upfront investment
   • CA pre-built AI service with a fixed set of algorithms
   • DAn open-source framework deployed on a cloud service with per-second billing
   Reveal Solution Discussion Select an answer first
2. 42application · medium

   An e-commerce company has multiple versions of a recommendation model in production. They need to roll back to a previous version if performance degrades, and they want to automate the deployment process. Which practice should they implement?

   • AStore all models in a shared network drive and overwrite the previous version
   • BManually copy model files to production servers whenever a new version is ready
   • CUse a model registry with versioning and a CI/CD pipeline that can deploy any registered version
   • DRetrain the model every week and deploy it without version tracking
   Reveal Solution Discussion Select an answer first
4. 43expert · hard

   A media company is building a data pipeline to train a content recommendation model. The pipeline must handle both batch and real-time data, and the model must be retrained frequently to reflect changing user preferences. The team has limited infrastructure budget and wants to avoid maintaining separate pipelines for batch and streaming. Which pipeline design is most cost-effective and maintainable?

   • AUse a unified pipeline that ingests both batch and streaming data into a common feature store, with scheduled retraining jobs
   • BBuild two separate pipelines: one for batch and one for streaming, and merge the outputs
   • CUse only a batch pipeline and ignore real-time data
   • DUse only a streaming pipeline and discard batch data
   Reveal Solution Discussion Select an answer first
5. 44foundation · easy

   What is the primary benefit of automating AI workflows using orchestration tools?

   • AGuaranteeing model accuracy
   • BEliminating the need for data governance
   • CIncreasing the size of training datasets
   • DReducing manual intervention and errors
   Reveal Solution Discussion Select an answer first
6. 45expert · hard

   A global bank is integrating an AI-based credit scoring model from a third-party vendor. The model will process customer data from multiple countries. The bank must comply with GDPR and local data residency laws, and the vendor must not retain any customer data. Which set of requirements should be included in the vendor contract?

   • AThe vendor is solely responsible for ensuring compliance with all local laws
   • BData processing in approved regions, no data retention by the vendor, and the right to audit the vendor's processing activities
   • CThe vendor may store data for up to 30 days for model improvement
   • DThe vendor can transfer data to any country for processing efficiency
   Reveal Solution Discussion Select an answer first