Verify Model Ready for Operationalization Go/no-Go Decision PMI-CPMAI Practice Questions (Page 3)

Questions 11–15

1. 11application · medium

   A team is evaluating a model that predicts equipment failure. The model was trained on data from machines manufactured by Company A. The team evaluates the model on a holdout set from Company A and achieves excellent performance. Before the go/no-go decision, they also evaluate on data from Company B's machines, which have different operating characteristics. The performance drops significantly. What is the most appropriate conclusion?

   • AThe model is ready for deployment because it performs well on the holdout set from Company A.
   • BThe model is ready for deployment because the drop in performance is expected when moving to a different domain.
   • CThe model should be deployed only for Company A's machines and not for Company B's machines.
   • DThe model is not ready for deployment because it does not generalize to data from Company B, which may be encountered in production.
   Reveal Solution Discussion Select an answer first
2. 12foundation · easy

   Why is it important to review deployment infrastructure before the go/no-go decision?

   • ATo ensure the model can be served with acceptable performance and reliability
   • BTo decide which features to include in the model
   • CTo check if the model's code is well documented
   • DTo determine the model's accuracy on the test set
   Reveal Solution Discussion Select an answer first
4. 13application · medium

   A medical imaging model meets its accuracy target on the standard test set. Before the go/no-go review, the team runs the model on images with varying brightness, contrast, and minor rotations to simulate real-world variability. The model's accuracy drops by 15% under these perturbations. What is the most appropriate conclusion?

   • AThe model is not robust enough for operationalization and should be retrained with data augmentation or similar techniques.
   • BThe model should be deployed but with a manual review for all images that fall outside the standard distribution.
   • CThe model is ready for deployment because it met the target on the standard test set.
   • DThe model is ready because the perturbations are artificial and will not occur in real clinical use.
   Reveal Solution Discussion Select an answer first
5. 14foundation · easy

   What is the basis for making the final go/no-go decision for model deployment?

   • AThe availability of a cloud budget
   • BThe opinion of the most senior data scientist
   • CAll evaluation results, readiness checks, and risk assessment
   • DThe model's performance on the training data
   Reveal Solution Discussion Select an answer first
6. 15foundation · easy

   Which activity best assesses model robustness?

   • AMeasuring the model's inference latency on a standard benchmark
   • BTesting the model on data with added noise and edge cases
   • CComparing the model's output to a human expert's judgment on a few samples
   • DReviewing the model's training code for syntax errors
   Reveal Solution Discussion Select an answer first