Describe the Solutions for Code, Data, and Model Traceability CERTIFIED-AI-EXPERT Practice Questions (Page 5)

Questions 21–25

1. 21application · medium

   A manufacturing company uses an AI model to predict equipment failure. The model is updated frequently, and the operations team needs to know which version of the predictive maintenance code was used for the model currently in production. What is the most reliable way to track this?

   • ARecord the Git commit hash of the code in the model's deployment metadata.
   • BStore the code in a shared folder with the date in the filename.
   • CKeep a log of when the model was last retrained.
   • DAsk the data science team to send an email when they update the code.
   Reveal Solution Discussion Select an answer first
2. 22foundation · easy

   Which tool is commonly used to track data versions and transformations in an AI pipeline?

   • ADVC (Data Version Control)
   • BDocker
   • CKubernetes
   • DTerraform
   Reveal Solution Discussion Select an answer first
4. 23application · medium

   A media company's recommendation model is retrained daily. The data science team wants to compare the performance of models trained on different days and understand which training data version each model used. What should they use?

   • AA file server with folders for each day's model and data.
   • BA backup system that stores daily snapshots of the model and data.
   • CA model registry that stores the model version, training run ID, and the dataset version.
   • DA spreadsheet that lists model names and training dates.
   Reveal Solution Discussion Select an answer first
5. 24expert · hard

   A pharmaceutical company must provide a complete audit trail for a drug-discovery model to meet regulatory compliance. The audit must show: (1) the exact code commit that generated the training features, (2) the raw data source and transformation steps, and (3) the model version and its training run. The team currently uses Git for code, a data lake for raw data, and a shared folder for models. They need a solution that integrates all three. What should they implement?

   • AA set of shell scripts that copy the model, code, and data into a timestamped archive folder after each training run.
   • BA Git repository that includes the model files and a copy of the raw data.
   • CA data lineage tool that only tracks the raw data source and its transformations.
   • DAn MLflow tracking server that logs parameters, metrics, and artifacts, and links each run to the Git commit and data source.
   Reveal Solution Discussion Select an answer first
6. 25expert · hard

   A bank is deploying a credit-scoring model and must comply with an audit requirement to explain the model's decisions. The audit team needs to trace a specific loan decision back to the model version, the training data version, and the feature engineering code. The team uses Git, a data warehouse, and a model registry, but these systems are not connected. What is the most effective approach to create the required end-to-end audit trail?

   • AUse a metadata management platform that ingests metadata from Git, the data warehouse, and the model registry, and correlates them via a shared run ID.
   • BCreate a manual process where the data science team documents the code, data, and model versions in a shared document after each deployment.
   • CStore all code, data, and models in a single Git repository.
   • DExport the model registry data to a spreadsheet and share it with the audit team.
   Reveal Solution Discussion Select an answer first