Kubeflow and ML Pipeline Orchestration Quiz

Answers and explanations appear as you take the quiz.

1. What underlying platform does Kubeflow run machine learning workflows on?
   • Apache Hadoop
   • A bare-metal Slurm cluster
   • Kubernetes
   • A single virtual machine
2. In Kubeflow Pipelines, how is each step of a workflow typically packaged?
   • As a containerized component running in a pod
   • As a stored procedure in a database
   • As a serverless function on a CDN
   • As a single monolithic script
3. What structure does a Kubeflow pipeline use to express the order and dependencies of steps?
   • A flat list executed top to bottom
   • A circular event loop
   • A relational schema
   • A directed acyclic graph (DAG)
4. Which Kubeflow component is dedicated to automated hyperparameter tuning?
   • KServe
   • Katib
   • MinIO
   • Argo
5. With KFP caching enabled, why might a rerun reuse a previous component result?
   • A matching eligible execution already has cached outputs
   • KFP never executes a component more than once
   • KFP ignores every input change
   • KFP randomly chooses which pods to skip
6. Which Kubeflow component is primarily used for serving trained models for inference?
   • Katib
   • Notebook servers
   • Argo CD
   • KServe
7. A team adopts KFP v2. What does the normal compiler output represent?
   • A trained model checkpoint
   • A PipelineSpec serialized as Intermediate Representation YAML
   • A universal Argo Workflows specification for every KFP version
   • A SQL query plan for the artifact database
8. A Kubeflow step reads a file overwritten at an unchanged URI. Why might its cache need to be disabled or the input version changed?
   • Kubernetes forbids external files
   • Every pipeline input is automatically content-hashed
   • The tracked input can stay the same while the file’s contents change
   • Caching always reruns components on every request