Determine the Differences Between Jupyter Notebooks and Pipelines CERTIFIED-AI-EXPERT Practice Questions (Page 5)

Questions 21–25

1. 21application · medium

   A data engineering team needs to build a data processing workflow that ingests data from multiple sources, cleans it, and loads it into a data warehouse. The workflow must run on a schedule and handle failures gracefully by retrying failed steps. Which approach is most appropriate?

   • AA pipeline with each step as a separate component, orchestrated by a workflow scheduler that supports retries.
   • BA single Python script that is run manually and logs errors to a file.
   • CA Jupyter notebook that is run manually by an engineer each time.
   • DA Jupyter notebook that is scheduled to run via a cron job.
   Reveal Solution Discussion Select an answer first
2. 22foundation · easy

   How does the execution model of a notebook differ from that of a pipeline?

   • ANotebooks are batch-oriented; pipelines are interactive.
   • BBoth use the same execution model but differ in user interface.
   • CNotebooks are stateless; pipelines maintain state across runs.
   • DNotebooks are stateful and interactive; pipelines are stateless and batch-oriented.
   Reveal Solution Discussion Select an answer first
4. 23expert · hard

   A team is building a model that requires a complex data preprocessing step that takes several hours. The team has prototyped the preprocessing in a Jupyter notebook. They need to run this preprocessing step as part of a nightly pipeline that also trains the model. The preprocessing step must be resumable if it fails. What is the most appropriate way to integrate the preprocessing logic?

   • AUse papermill to execute the notebook as a step in the pipeline.
   • BConvert the preprocessing logic into a single monolithic script and run it as a step in the pipeline.
   • CRun the preprocessing in the notebook and save the output to a file, then have the pipeline read that file.
   • DRefactor the preprocessing logic into a pipeline step that saves intermediate results to disk and can be resumed from the last checkpoint.
   Reveal Solution Discussion Select an answer first
5. 24application · medium

   A machine learning engineer needs to deploy a model that processes a high volume of incoming requests. The inference logic was prototyped in a Jupyter notebook. The team must ensure the service scales horizontally and restarts cleanly if a node fails. What is the most appropriate approach?

   • ASchedule the notebook to run every minute to process the incoming requests in batches.
   • BRun the notebook on a large VM and expose the kernel via a REST API for the incoming requests.
   • CRefactor the notebook logic into a stateless pipeline or service module and deploy it as a containerized microservice.
   • DConvert the notebook to a Python script and run it as a long-lived process on a single server.
   Reveal Solution Discussion Select an answer first
6. 25application · medium

   A company needs to deploy a machine learning model that will be used by a batch job that runs every night. The batch job processes a large dataset and produces predictions. The model was developed in a Jupyter notebook. What is the most appropriate way to deploy the model for this batch job?

   • ARefactor the model inference logic into a pipeline step that can be invoked by the nightly batch job.
   • BDeploy the notebook as a web service using a tool like Voilà.
   • CConvert the notebook to a Python script and run it manually each night.
   • DRun the notebook manually each night to generate the predictions.
   Reveal Solution Discussion Select an answer first