5.3 Explain AI Workflow Design and Monitoring Principles AI-TECHNICAL-PRACTITIONER Practice Questions (Page 4)

Questions 16–20

1. 16application · medium

   A media company runs an AI model that recommends articles to users. The model is retrained weekly. The company wants to monitor the model's performance in production. Which metric is most appropriate to track?

   • AThe number of features used in the model.
   • BThe model's training loss on the last training dataset.
   • CThe time taken to retrain the model each week.
   • DUser engagement metrics such as click-through rate and time spent on recommended articles.
   Reveal Solution Discussion Select an answer first
2. 17foundation · easy

   Which monitoring principle focuses on ensuring that the infrastructure running an AI workflow has sufficient CPU, memory, and storage?

   • AMonitoring data drift
   • BChecking system health
   • CTracking model performance
   • DReviewing model version
   Reveal Solution Discussion Select an answer first
4. 18application · medium

   A manufacturing company has an AI workflow that predicts equipment failure. The workflow has multiple stages: data ingestion from sensors, preprocessing, model training, and alert generation. The company wants to ensure that if the preprocessing stage fails, the alert generation stage does not run. What is the best way to achieve this?

   • AUse a timer to trigger the alert generation stage regardless of preprocessing status.
   • BHave the alert generation stage check for the existence of preprocessed data and skip if missing.
   • CRun all stages in parallel and let the alert generation stage handle missing data.
   • DConfigure the orchestration tool to run the alert generation stage only after the preprocessing stage completes successfully.
   Reveal Solution Discussion Select an answer first
5. 19expert · hard

   A streaming service uses an AI model to recommend content. The model is retrained monthly using historical viewing data. Recently, the service noticed a decline in user engagement. The data team suspects that the model is not adapting quickly enough to changing user preferences. The team wants to implement a feedback loop that responds faster to trends. Which approach is most effective?

   • AContinue with monthly retraining but increase the amount of historical data used.
   • BManually adjust the model's recommendations based on current popular content.
   • CImplement a feedback loop that collects real-time user interactions and triggers retraining when significant changes are detected.
   • DFreeze the model to ensure consistent recommendations.
   Reveal Solution Discussion Select an answer first
6. 20foundation · easy

   How does a feedback loop improve an AI model over time?

   • ABy permanently freezing the model to avoid changes
   • BBy using model predictions and outcomes to retrain or fine-tune the model
   • CBy increasing the model's complexity without new data
   • DBy deleting all historical data to reduce storage
   Reveal Solution Discussion Select an answer first