Tracking and Running ML Experiments PROFESSIONAL-MACHINE-LEARNING-ENGINEER Practice Questions (Page 4)

Questions 16–20

1. 16foundation · easy

   What is the primary function of Gemini Enterprise Agent Platform Pipelines?

   • ATo store and version ML models
   • BTo monitor model performance in production
   • CTo orchestrate and manage ML pipelines for agent-based experimentation
   • DTo provide a Jupyter notebook environment
   Reveal Solution Discussion Select an answer first
2. 17application · medium

   A team is developing a generative AI model that produces product descriptions for an e-commerce site. They need to evaluate the quality of the descriptions, including factual accuracy, tone, and adherence to brand guidelines. They have a small set of human-annotated examples. Which evaluation approach is most appropriate?

   • AUse a combination of human evaluation and LLM-as-a-judge to assess qualitative aspects
   • BUse a simple keyword matching approach to check if important product features are mentioned
   • CUse only automated metrics like BLEU and ROUGE to compare generated text to reference descriptions
   • DUse only perplexity as a metric to evaluate the fluency of the generated text
   Reveal Solution Discussion Select an answer first
4. 18foundation · easy

   Which tool is used to store and manage metadata for ML experiments on Google Cloud?

   • ACloud Functions
   • BBigQuery
   • CCloud Storage
   • DGemini Enterprise Agent Platform ML Metadata
   Reveal Solution Discussion Select an answer first
5. 19application · medium

   A machine learning engineer is building a Kubeflow Pipelines pipeline to train and evaluate a model. They need to pass the trained model artifact from the training step to the evaluation step, and also record the model's hyperparameters for lineage tracking. What is the correct way to define the component interfaces?

   • AUse a global variable in the pipeline code to hold the model object
   • BSerialize the model to a base64 string and pass it as a string parameter
   • CUse a shared file path in a persistent volume to pass the model between steps
   • DDefine the training component to output a Model artifact and the evaluation component to accept a Model artifact as input
   Reveal Solution Discussion Select an answer first
6. 20foundation · easy

   What is the purpose of compiling a Kubeflow pipeline?

   • ATo convert the pipeline code into a format that can be executed by the Kubeflow Pipelines engine
   • BTo test the pipeline on sample data
   • CTo compress the pipeline code to reduce storage
   • DTo encrypt the pipeline code for security
   Reveal Solution Discussion Select an answer first