5.3 Explain AI Workflow Design and Monitoring Principles AI-TECHNICAL-PRACTITIONER Practice Questions (Page 1)

Questions 1–5

1. 1foundation · easy

   Which ethical principle in AI workflow design is primarily concerned with avoiding bias against certain groups of people?

   • AEfficiency
   • BFairness
   • CAccountability
   • DTransparency
   Reveal Solution Discussion Select an answer first
2. 2application · medium

   A healthcare startup is building an AI workflow to predict patient readmission risk. The data science team wants to experiment with different feature engineering approaches, while the operations team requires that the model be retrained automatically when new patient data arrives. The company has a strict policy that all data processing must remain within their own cloud subscription. Which design should the team adopt?

   • ABuild separate stages for ingestion, preprocessing, training, and deployment, orchestrated by a workflow tool that triggers the next stage only when the previous one succeeds.
   • BRun the entire workflow on a serverless function that executes every time a new file is uploaded to cloud storage, with all stages in a single function.
   • CTrain the model on-premises and use a cloud-based API for deployment, with no automated retraining.
   • DUse a single monolithic pipeline that ingests raw data, trains the model, and deploys it, with the entire pipeline triggered by a schedule.
   Reveal Solution Discussion Select an answer first
4. 3foundation · easy

   Which governance principle ensures that the decisions made by an AI model can be understood and explained to stakeholders?

   • AScalability
   • BAccountability
   • CTransparency
   • DFairness
   Reveal Solution Discussion Select an answer first
5. 4expert · hard

   A financial institution is building an AI workflow for credit scoring. The workflow includes data ingestion, feature engineering, model training, and deployment. The institution has a strict requirement that the model must be retrained whenever new regulatory rules are introduced. The data science team wants to experiment with new features without affecting the production model. The operations team wants to minimize downtime during model updates. Which approach best meets all requirements?

   • AFreeze the model and only update it annually, regardless of regulatory changes.
   • BUse a CI/CD pipeline for the AI workflow, with separate environments for development, testing, and production, and automated retraining triggers.
   • CUse a manual process where data scientists update the model and then deploy it directly.
   • DRetrain the model in production whenever new regulations are introduced, without a separate testing environment.
   Reveal Solution Discussion Select an answer first
6. 5application · medium

   An online education platform uses an AI model to recommend courses to students. The platform wants to improve the model based on student feedback and outcomes. Which approach best implements a feedback loop?

   • AUse a static set of popular courses as recommendations, updated annually.
   • BMonitor the model's accuracy on a held-out test set and retrain only when accuracy drops.
   • CAsk a panel of experts to manually adjust the recommendation algorithm each month.
   • DCollect data on which courses students complete and their satisfaction ratings, then use this data to retrain the model.
   Reveal Solution Discussion Select an answer first