1.4 AI/Machine Learning 400-007 Practice Questions (Page 7)

Questions 31–35

1. 31application · medium

   A ride-sharing company runs an AI model for demand prediction. The model inference workload spikes during rush hours and holidays, but is low at other times. The company wants to minimize cost while ensuring that inference latency remains below 100 ms during peak times. Which auto-scaling strategy should the company adopt?

   • AImplement horizontal auto-scaling based on CPU utilization, with a minimum of one instance and a maximum of ten.
   • BUse a serverless inference service that automatically scales to zero during idle periods and scales up during spikes.
   • CProvision a fixed number of inference instances to handle peak load, ensuring consistent performance.
   • DUse a scheduled auto-scaling policy that increases capacity during known peak hours and reduces it during off-peak hours.
   Reveal Solution Discussion Select an answer first
2. 32foundation · easy

   Why is data quality important for maintaining data integrity in AI/ML?

   • ABecause data quality is required to meet network security standards
   • BBecause data quality determines the speed of model training
   • CBecause data quality reduces the need for data storage
   • DBecause poor data quality can lead to biased or inaccurate model predictions
   Reveal Solution Discussion Select an answer first
4. 33foundation · easy

   What does data sovereignty primarily refer to in the context of AI/ML workloads?

   • AThe ownership of data by the organization that collects it
   • BThe technical ability to replicate data across multiple cloud regions for high availability
   • CThe legal and regulatory requirement that data remain within specific geographical boundaries
   • DThe encryption of data both at rest and in transit
   Reveal Solution Discussion Select an answer first
5. 34expert · hard

   A video streaming service is deploying an AI model for content recommendation. The model is trained on user viewing history and is updated daily. The training data is stored in a data lake, and the inference service is deployed on a Kubernetes cluster. The service experiences high traffic during evening hours. The company wants to minimize the cost of data transfer between the data lake and the training cluster. Which design should the company adopt?

   • AUse a VPN to connect the data lake and training cluster securely, regardless of location.
   • BUse a content delivery network (CDN) to cache the training data for the training cluster.
   • CMove the training cluster to the same cloud region as the data lake to reduce data transfer costs.
   • DStore the training data on the inference cluster to avoid transfer altogether.
   Reveal Solution Discussion Select an answer first
6. 35foundation · easy

   Which cost is typically considered a recurring operational cost in an AI/ML deployment?

   • AInitial model development and training
   • BCloud compute resources used for inference
   • CPurchase of training hardware
   • DOne-time data labeling costs
   Reveal Solution Discussion Select an answer first