DevOps and MLOps DY0-001 Practice Questions (Page 1)

Questions 1–5

1. 1application · medium

   A fintech company has a mature CI pipeline for its fraud-detection model. Every merge to main triggers automated tests, and the model is packaged as a container image. The team now wants to automatically push the new model to production as soon as it passes all tests, with no human approval step. Which approach best achieves this while maintaining the ability to detect problems quickly?

   • ADeploy the container image to a staging environment first, and have a human manually promote it to production after a review.
   • BAutomatically deploy the container image to production only on weekends, when transaction volume is lower.
   • CSkip the container image and have the CI pipeline directly replace the model files on the production server.
   • DConfigure the CI pipeline to automatically deploy the container image to the production environment, and rely on performance monitoring dashboards to detect issues after release.
   Reveal Solution Discussion Select an answer first
2. 2application · medium

   A data science team has deployed a sentiment-analysis model. The model's accuracy on a test set was 92%. In production, the team monitors the model's predictions and notices that the distribution of predicted sentiment scores has shifted significantly from the distribution seen during validation. The model's code and input features have not changed. What is the most appropriate action to take?

   • ARoll back to a previous version of the model, as the current version is likely buggy.
   • BIgnore the shift, as the model's accuracy on the test set was high.
   • CInvestigate the input data for drift, as the shift in prediction distribution suggests the model is seeing data it was not trained on.
   • DRetrain the model immediately on the same training data to improve its performance.
   Reveal Solution Discussion Select an answer first
4. 3application · medium

   A media company deploys a content-recommendation model as a REST API. The model was trained on user interaction data. After a major UI redesign, the product team notices that user click patterns have changed, but the model's input features (e.g., time on page, scroll depth) are still being collected in the same format. The model's performance metrics (e.g., click-through rate) have declined. What should the team monitor to detect this issue early?

   • ACheck the model's training loss on the original training dataset.
   • BReview the model's code for any recent changes that might have introduced a bug.
   • CTrack the distribution of the input features over time and compare them to the training data distribution.
   • DMonitor the model's prediction latency to ensure the API is responding quickly.
   Reveal Solution Discussion Select an answer first
5. 4foundation · easy

   Which practice is a core principle of continuous integration (CI)?

   • ADevelopers work on long-lived feature branches and merge them only after the project is complete
   • BManual code reviews are the only way to ensure code quality before merging
   • CAutomated builds and tests run on every commit to the main branch
   • DDeployments to production are performed manually by operations staff after each release
   Reveal Solution Discussion Select an answer first
6. 5expert · hard

   A financial services company has a model that predicts credit risk. The model is retrained monthly and deployed via a CI/CD pipeline. The team monitors the model's AUC (Area Under the ROC Curve) on a rolling basis. After a recent deployment, the AUC dropped from 0.85 to 0.80. The team suspects the issue is with the new model, not the data. Which action should they take first to confirm this?

   • ARoll back to the previous model immediately.
   • BCompare the new model's predictions on a fixed validation dataset against the previous model's predictions.
   • CCheck the input data for drift by comparing the current production data distribution to the training data.
   • DRetrain the model on the same training data and redeploy it.
   Reveal Solution Discussion Select an answer first