Developing End-To-End ML Pipelines PROFESSIONAL-MACHINE-LEARNING-ENGINEER Practice Questions (Page 5)

Questions 21–25

1. 21expert · hard

   A gaming company needs to build a real-time ML pipeline that ingests player events from Pub/Sub, performs feature engineering, and triggers model retraining when the data distribution shifts. They have a team of ML engineers who are comfortable with Kubernetes and want to avoid vendor lock-in. They also need to handle bursty traffic and scale the pipeline components independently. Which solution best meets these requirements?

   • AUse a single Ray cluster on GKE to handle both stream processing and model retraining.
   • BUse Vertex AI Pipelines with a Pub/Sub trigger to run the pipeline on a schedule.
   • CBuild a custom pipeline using Apache Beam on Dataflow for stream processing and trigger retraining with a custom Kubernetes operator.
   • DUse Cloud Composer with a Pub/Sub sensor to trigger the pipeline steps.
   Reveal Solution Discussion Select an answer first
2. 22foundation · easy

   What is the main benefit of using pre-built pipeline templates in ML development?

   • AThey eliminate the need for any custom code in your pipeline
   • BThey automatically deploy your model to production
   • CThey guarantee that your model will achieve high accuracy
   • DThey accelerate development by providing reusable, tested components and best practices
   Reveal Solution Discussion Select an answer first
4. 23application · medium

   A company is starting a new ML project and wants to use a pre-built pipeline template from Vertex AI Pipelines. The template includes data validation, training, and evaluation components. What is the main advantage of using this template?

   • AIt requires no customization for the specific use case.
   • BIt eliminates the need for data validation.
   • CIt accelerates development by providing a working pipeline with best practices.
   • DIt guarantees the model will have high accuracy.
   Reveal Solution Discussion Select an answer first
5. 24expert · hard

   A large enterprise has a complex ML pipeline with multiple teams contributing different steps. They need to orchestrate the pipeline, but they have strict data residency requirements that prevent them from using a fully managed service that stores pipeline metadata outside their region. They have a strong Kubernetes team and want to use an open-source tool. Which approach is most appropriate?

   • AUse Vertex AI Pipelines with the pipeline metadata stored in a regional bucket.
   • BUse a custom Python script with Apache Airflow on a Compute Engine VM.
   • CDeploy Kubeflow Pipelines on their own GKE cluster in the required region.
   • DUse Cloud Composer with a private IP configuration.
   Reveal Solution Discussion Select an answer first
6. 25foundation · easy

   In Google Cloud's Vertex AI, what is a common way to use pre-built pipeline templates?

   • ABy using only templates that are automatically generated from your data
   • BBy writing a new pipeline from scratch every time
   • CBy cloning and customizing a template from the Vertex AI Pipelines template gallery
   • DBy deploying templates directly without any modification
   Reveal Solution Discussion Select an answer first