A company is building the operational framework for an AI data infrastructure. Which TWO of the following are activities that are routinely performed to maintain healthy data governance, while being distinct from model development or infrastructure operations tasks? (Choose TWO.)
Select two data governance activities.
Metadata management — attaching and maintaining descriptions and attribute information to each data item
This is correct. Attaching and maintaining metadata such as meaning, attributes, and update dates to each data item is a governance activity that supports understanding and appropriate handling of data.
Data versioning — keeping a change history of the data
This is correct. Keeping a version history of changes to datasets is a governance activity that supports reproducibility and issue tracking.
Reviewing the scaling configuration of inference endpoints
This is incorrect. Reviewing scaling configurations is an infrastructure operations task. While it is necessary for operating AI systems, it is not a data governance activity whose subject is the management and control of data.
Improving prompt templates
This is incorrect. Improving prompt templates is a prompt engineering task for improving output quality. It is not a data governance activity whose subject is the management and control of data.
Reducing costs by purchasing reserved GPU instances
This is incorrect. Purchasing reserved instances is a cost optimization effort. It is not a data governance activity whose subject is the management and control of data.
Data governance activities include metadata management (attaching and maintaining descriptions and attributes to each data item) and data versioning (keeping a change history), among others such as lineage, cataloging, classification, and access management. These activities support understanding, reproducibility, and control.