Verify Model Ready for Operationalization Go/no-Go Decision PMI-CPMAI Practice Questions (Page 4)

Questions 16–20

1. 16application · medium

   A team plans to deploy a real-time recommendation model that currently runs on a GPU server with 32 GB of memory. The production environment will use a CPU-only server with 16 GB of memory. The model's inference latency on the CPU server is 800 ms, but the service-level agreement (SLA) requires a maximum latency of 500 ms. What should the team do before the go/no-go decision?

   • AIssue a no-go decision and recommend either upgrading the production infrastructure or optimizing the model to meet the latency requirement.
   • BReduce the model's batch size to lower memory usage and assume latency will improve.
   • CDeploy the model and rely on horizontal scaling to reduce latency below 500 ms.
   • DProceed with deployment because the model meets the accuracy targets and the SLA can be adjusted later.
   Reveal Solution Discussion Select an answer first
2. 17application · medium

   A fraud-detection team trained a model to identify suspicious transactions. The predefined success criteria require a recall of at least 0.92 on the holdout test set. During final evaluation, the model achieves a recall of 0.90 but a precision of 0.97, which exceeds the precision target. The team lead is under pressure to deploy because the precision is excellent. What should the team do?

   • ADeploy the model because precision exceeds the target and false positives are costly.
   • BIssue a no-go decision because the recall target was not met, and investigate retraining or threshold adjustment.
   • CRe-run the evaluation on a different holdout set to see if recall improves above 0.92.
   • DDeploy the model but monitor recall closely for the first month to see if it improves in production.
   Reveal Solution Discussion Select an answer first
4. 18foundation · easy

   During the go/no-go evaluation, a data scientist compares the model's F1 score, precision, and recall on the validation set to the thresholds defined in the project charter. What is the primary purpose of this activity?

   • ATo identify which features contribute most to the model's predictions
   • BTo verify that the model produces the same output for identical inputs
   • CTo estimate the cost of running the model in production
   • DTo determine whether the model meets the predefined success criteria
   Reveal Solution Discussion Select an answer first
5. 19foundation · easy

   Which dataset is most appropriate for verifying a model's generalization to unseen data?

   • AA validation set used for early stopping during training
   • BThe training dataset, to confirm the model has learned the patterns
   • CA synthetic dataset generated from the training data distribution
   • DA holdout test set that was not used during training or hyperparameter tuning
   Reveal Solution Discussion Select an answer first
6. 20foundation · easy

   Which of the following is an example of essential content in model documentation for operationalization?

   • AThe project manager's personal notes
   • BThe marketing plan for the AI product
   • CThe team's meeting minutes
   • DThe model's input and output schema
   Reveal Solution Discussion Select an answer first