Question 330
At fictional Seabrook Industrial Services, a Snowflake data and AI engineering team is conducting a regulated customer response automation exercise with legal, privacy and information-security review for any restricted inputs. The separately composed technical incident is: A data science team registers a new inference model but deploys an endpoint with a different runtime dependency version from the tested environment, causing inconsistent output fields. Assume only these documented facts and use safe, authorized production-change procedures; exact current Snowflake feature syntax and region or account availability require current review. Which repeatable assurance or control is most relevant to preventing this situation?
Explanation
Writer-proposed answer: B (UNVERIFIED)
Preliminary individual option explanations — NOT expert validated
Correct Answer
Require build manifests, deployment smoke tests, artifact-to-endpoint checks and controlled dependency updates. Model correctness depends on both the registered artifact and its serving environment; reproducible deployment must cover both.
Incorrect Answers
This response concerns Embedding dimension mismatch in a vector column, which is different from the stated failure. A data science team registers a new inference model but deploys an endpoint with a different runtime dependency version from the tested environment, causing inconsistent output fields.
Verified query repository for business metric questions is not the cause of the reported failure: A data science team registers a new inference model but deploys an endpoint with a different runtime dependency version from the tested environment, causing inconsistent output fields.
Cortex Search incremental refresh lag alone does not establish the needed evidence: Replayed acceptance inputs produce matching validated outputs in staging and production, and dependency hashes match the approved release.