AI Project Life Cycle, MLOps, and DataOps CAIPM Practice Questions (Page 9)

Questions 41–45

1. 41application · medium

   An e-commerce company is building a recommendation system. The data engineering team manages the feature pipeline, and the ML team manages model training and deployment. The program manager notices that feature changes often break the model in production because the teams work in silos. Which approach BEST integrates MLOps and DataOps to reduce these failures?

   • AMove all feature engineering into the model training code.
   • BEstablish a shared feature store with versioned features and joint CI/CD pipelines.
   • CHave the data engineering team test the model after every feature change.
   • DAsk the ML team to freeze the model inputs to avoid future changes.
   Reveal Solution Discussion Select an answer first
2. 42application · medium

   A retail company is starting an AI project to forecast inventory demand. The data science team has built a prototype model in a notebook that shows promising accuracy on historical data. The project sponsor wants to move to production quickly. Which action should the program manager take FIRST to align with the AI project life cycle?

   • ASet up a CI/CD pipeline for the model code to enable rapid updates.
   • BAsk the data engineering team to build a new data pipeline for the model's input features.
   • CDeploy the notebook model to a production endpoint and monitor its performance.
   • DReview the problem definition and success criteria with stakeholders before proceeding.
   Reveal Solution Discussion Select an answer first
4. 43application · medium

   A startup is building a recommendation engine and wants to implement MLOps from the start. The team is small and has limited infrastructure. They need a tool that can version both code and data, and also track model experiments. Which tool would best meet these needs?

   • AApache Airflow
   • BKubernetes
   • CMLflow
   • DGit
   Reveal Solution Discussion Select an answer first
5. 44application · medium

   A financial services firm has a credit-risk model in production. The model's performance degrades gradually as customer behavior changes. The ML team currently retrains the model manually every quarter. Which MLOps practice would BEST address the gradual performance degradation?

   • AContinuous delivery of the model to a staging environment only.
   • BContinuous integration of code changes into the main branch.
   • CVersion control for the training data and model artifacts.
   • DContinuous training with automated retraining triggers based on performance drift.
   Reveal Solution Discussion Select an answer first
6. 45foundation · easy

   Which sequence correctly represents the typical phases of an AI project life cycle, from start to finish?

   • AModel development, data preparation, problem definition, deployment, retirement
   • BProblem definition, data preparation, model development, deployment, maintenance
   • CData collection, model training, problem definition, deployment, monitoring
   • DDeployment, model development, data preparation, problem definition, maintenance
   Reveal Solution Discussion Select an answer first