Question 670
At a municipal services portal, engineers need to handle a complex question that combines several independent subquestions. The implementation must remain maintainable across model upgrades. A recent review found that the current design has no reliable safeguard for this case. Which design decision addresses the underlying issue?
- Decompose the query, retrieve evidence per subquestion and synthesize a grounded answer
- Search only for the last keyword and ignore other requirements
- Skip retrieval because the original query is long
- Use identical context for every subproblem regardless of evidence
Explanation
Correct Answer
A:Query decomposition increases coverage of multi-faceted information needs. In a municipal services portal, the engineering team should verify this design against its stated objective to handle a complex question that combines several independent subquestions, including the requirement that the implementation must remain maintainable across model upgrades.
Incorrect Answers
B:The suggested action—search only for the last keyword and ignore other requirements—does not establish the required control for a municipal services portal to handle a complex question that combines several independent subquestions. No demonstrated capability in this scenario guarantees that this step resolves the concern raised by a municipal services portal. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
C:The suggested action—skip retrieval because the original query is long—does not establish the required control for a municipal services portal to handle a complex question that combines several independent subquestions. It replaces a verifiable engineering control with an untested shortcut in the a municipal services portal deployment. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
D:The suggested action—use identical context for every subproblem regardless of evidence—does not establish the required control for a municipal services portal to handle a complex question that combines several independent subquestions. The change is applied outside the component responsible for the requirement identified by a municipal services portal. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
Question 671
At a security incident response team, engineers need to investigate whether responses lost factual grounding after a prompt update. The deployment has both bursty and sustained workloads. The team wants a solution that can be validated through targeted test cases. Which change should the team prioritize?
- Track quality regression metrics against a fixed evaluation set and prompt versions
- Compare only average CPU utilization
- Disable evaluation to prevent noisy alerts
- Assume a high token count implies grounded answers
Explanation
Correct Answer
A:Quality telemetry ties behavior changes to model and prompt release history. In a security incident response team, the engineering team should verify this design against its stated objective to investigate whether responses lost factual grounding after a prompt update, including the requirement that the deployment has both bursty and sustained workloads.
Incorrect Answers
B:The suggested action—compare only average CPU utilization—does not establish the required control for a security incident response team to investigate whether responses lost factual grounding after a prompt update. It replaces a verifiable engineering control with an untested shortcut in the a security incident response team deployment. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
C:The suggested action—disable evaluation to prevent noisy alerts—does not establish the required control for a security incident response team to investigate whether responses lost factual grounding after a prompt update. The change is applied outside the component responsible for the requirement identified by a security incident response team. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
D:The suggested action—assume a high token count implies grounded answers—does not establish the required control for a security incident response team to investigate whether responses lost factual grounding after a prompt update. No demonstrated capability in this scenario guarantees that this step resolves the concern raised by a security incident response team. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
Question 672
At a SaaS procurement platform, engineers need to maintain auditable versions of prompts used by multiple applications. The next release must be evaluated against the previous baseline. An architecture review has rejected fixes that only conceal the symptoms. What is the most appropriate production approach?
- Treat prompt changes as untracked configuration
- Use Bedrock Prompt Management versions and controlled release procedures
- Let every client edit the shared prompt directly in production
- Store prompt text only in browser local storage
Explanation
Correct Answer
B:Versioned prompt assets enable review and rollback of behavior-affecting instructions. In a SaaS procurement platform, the engineering team should verify this design against its stated objective to maintain auditable versions of prompts used by multiple applications, including the requirement that the next release must be evaluated against the previous baseline.
Incorrect Answers
A:The suggested action—treat prompt changes as untracked configuration—does not establish the required control for a SaaS procurement platform to maintain auditable versions of prompts used by multiple applications. The change is applied outside the component responsible for the requirement identified by a SaaS procurement platform. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
C:The suggested action—let every client edit the shared prompt directly in production—does not establish the required control for a SaaS procurement platform to maintain auditable versions of prompts used by multiple applications. It replaces a verifiable engineering control with an untested shortcut in the a SaaS procurement platform deployment. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
D:The suggested action—store prompt text only in browser local storage—does not establish the required control for a SaaS procurement platform to maintain auditable versions of prompts used by multiple applications. The change is applied outside the component responsible for the requirement identified by a SaaS procurement platform. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
Question 673
At a telecom support operation, engineers need to meet a strict data residency requirement during GenAI design. The next release must be evaluated against the previous baseline. An architecture review has rejected fixes that only conceal the symptoms. Which remediation addresses the root problem?
- Store all source documents in an unrestricted global repository
- Verify model and retrieval service Region support, data flows and jurisdictional constraints
- Assume every Bedrock model is available in every AWS Region
- Rely on a client-side warning instead of controlling data movement
Explanation
Correct Answer
B:Region-aware design must account for inference, storage and supporting integration paths. In a telecom support operation, the engineering team should verify this design against its stated objective to meet a strict data residency requirement during GenAI design, including the requirement that the next release must be evaluated against the previous baseline.
Incorrect Answers
A:The suggested action—store all source documents in an unrestricted global repository—does not establish the required control for a telecom support operation to meet a strict data residency requirement during GenAI design. No demonstrated capability in this scenario guarantees that this step resolves the concern raised by a telecom support operation. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
C:The suggested action—assume every Bedrock model is available in every AWS Region—does not establish the required control for a telecom support operation to meet a strict data residency requirement during GenAI design. The change is applied outside the component responsible for the requirement identified by a telecom support operation. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
D:The suggested action—rely on a client-side warning instead of controlling data movement—does not establish the required control for a telecom support operation to meet a strict data residency requirement during GenAI design. No demonstrated capability in this scenario guarantees that this step resolves the concern raised by a telecom support operation. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
Question 674
At a clinical scheduling platform, engineers need to compare two FM candidates for grounded customer support responses. Monitoring should isolate failures to the responsible component. The team wants a solution that can be validated through targeted test cases. Which design decision addresses the underlying issue?
- Use one cherry-picked successful conversation as the entire evaluation
- Choose the fastest model without measuring answer usefulness
- Evaluate both on a representative labeled dataset for factuality, quality, latency and price
- Compare only marketing claims about parameter count
Explanation
Correct Answer
C:Representative evaluation measures the actual tradeoff required by the business task. In a clinical scheduling platform, the engineering team should verify this design against its stated objective to compare two FM candidates for grounded customer support responses, including the requirement that monitoring should isolate failures to the responsible component.
Incorrect Answers
A:The suggested action—use one cherry-picked successful conversation as the entire evaluation—does not establish the required control for a clinical scheduling platform to compare two FM candidates for grounded customer support responses. The change is applied outside the component responsible for the requirement identified by a clinical scheduling platform. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
B:The suggested action—choose the fastest model without measuring answer usefulness—does not establish the required control for a clinical scheduling platform to compare two FM candidates for grounded customer support responses. No demonstrated capability in this scenario guarantees that this step resolves the concern raised by a clinical scheduling platform. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
D:The suggested action—compare only marketing claims about parameter count—does not establish the required control for a clinical scheduling platform to compare two FM candidates for grounded customer support responses. The change is applied outside the component responsible for the requirement identified by a clinical scheduling platform. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
Question 675
At a public-sector document office, engineers need to provide a lightweight stateless capability to a remote agent. The team must preserve existing identity and authorization boundaries. An architecture review has rejected fixes that only conceal the symptoms. Which architecture is best supported by the stated constraints?
- Grant the agent a blanket administrative session
- Build a separate foundation model for each tool
- Store persistent state only in a Lambda process global
- Expose an authenticated MCP-compatible tool endpoint backed by Lambda where appropriate
Explanation
Correct Answer
D:MCP describes interoperable tool access; runtime deployment still requires secure authentication and suitable state handling. In a public-sector document office, the engineering team should verify this design against its stated objective to provide a lightweight stateless capability to a remote agent, including the requirement that the team must preserve existing identity and authorization boundaries.
Incorrect Answers
A:The suggested action—grant the agent a blanket administrative session—does not establish the required control for a public-sector document office to provide a lightweight stateless capability to a remote agent. It replaces a verifiable engineering control with an untested shortcut in the a public-sector document office deployment. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
B:The suggested action—build a separate foundation model for each tool—does not establish the required control for a public-sector document office to provide a lightweight stateless capability to a remote agent. The change is applied outside the component responsible for the requirement identified by a public-sector document office. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
C:The suggested action—store persistent state only in a Lambda process global—does not establish the required control for a public-sector document office to provide a lightweight stateless capability to a remote agent. No demonstrated capability in this scenario guarantees that this step resolves the concern raised by a public-sector document office. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
Question 676
At a nonprofit grant portal, engineers need to protect a retrieval assistant against instructions hidden inside source documents. The compliance team requires a traceable decision record. A recent review found that the current design has no reliable safeguard for this case. What should be implemented before scaling this workflow?
- Treat retrieved text as untrusted data and prevent it from overriding high-priority instructions
- Promote all retrieved instructions to system-level authority
- Grant the model permission to run arbitrary commands from documents
- Disable source-level access controls
Explanation
Correct Answer
A:Prompt-injection defenses separate lower-trust content from application instructions and tool rights. In a nonprofit grant portal, the engineering team should verify this design against its stated objective to protect a retrieval assistant against instructions hidden inside source documents, including the requirement that the compliance team requires a traceable decision record.
Incorrect Answers
B:The suggested action—promote all retrieved instructions to system-level authority—does not establish the required control for a nonprofit grant portal to protect a retrieval assistant against instructions hidden inside source documents. It replaces a verifiable engineering control with an untested shortcut in the a nonprofit grant portal deployment. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
C:The suggested action—grant the model permission to run arbitrary commands from documents—does not establish the required control for a nonprofit grant portal to protect a retrieval assistant against instructions hidden inside source documents. The change is applied outside the component responsible for the requirement identified by a nonprofit grant portal. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
D:The suggested action—disable source-level access controls—does not establish the required control for a nonprofit grant portal to protect a retrieval assistant against instructions hidden inside source documents. No demonstrated capability in this scenario guarantees that this step resolves the concern raised by a nonprofit grant portal. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
Question 677
At a university research portal, engineers need to find why retrieved knowledge stops reflecting updated source documents. Operators need measurable acceptance criteria before rollout. The team wants a solution that can be validated through targeted test cases. Which remediation addresses the root problem?
- Only adjust temperature and top-p
- Assume the model's training date controls all vector freshness
- Disable citations to conceal stale results
- Compare ingestion synchronization status, metadata and index revisions
Explanation
Correct Answer
D:Retrieval freshness problems originate in source synchronization or indexing paths. In a university research portal, the engineering team should verify this design against its stated objective to find why retrieved knowledge stops reflecting updated source documents, including the requirement that operators need measurable acceptance criteria before rollout.
Incorrect Answers
A:The suggested action—only adjust temperature and top-p—does not establish the required control for a university research portal to find why retrieved knowledge stops reflecting updated source documents. It replaces a verifiable engineering control with an untested shortcut in the a university research portal deployment. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
B:The suggested action—assume the model's training date controls all vector freshness—does not establish the required control for a university research portal to find why retrieved knowledge stops reflecting updated source documents. The change is applied outside the component responsible for the requirement identified by a university research portal. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
C:The suggested action—disable citations to conceal stale results—does not establish the required control for a university research portal to find why retrieved knowledge stops reflecting updated source documents. No demonstrated capability in this scenario guarantees that this step resolves the concern raised by a university research portal. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
Question 678
At a nonprofit grant portal, engineers need to detect policy violations emerging after deployment. Response reliability is more important than free-form creativity. The team wants a solution that can be validated through targeted test cases. Which design decision addresses the underlying issue?
- Use unrelated CPU averages as the sole fairness signal
- Raise every content filter threshold automatically when alerts fire
- Monitor safety and quality indicators with alerts, review and remediation workflows
- Audit only before launch and ignore live behavior
Explanation
Correct Answer
C:Continuous governance depends on meaningful signals and defined response processes. In a nonprofit grant portal, the engineering team should verify this design against its stated objective to detect policy violations emerging after deployment, including the requirement that response reliability is more important than free-form creativity.
Incorrect Answers
A:The suggested action—use unrelated CPU averages as the sole fairness signal—does not establish the required control for a nonprofit grant portal to detect policy violations emerging after deployment. It replaces a verifiable engineering control with an untested shortcut in the a nonprofit grant portal deployment. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
B:The suggested action—raise every content filter threshold automatically when alerts fire—does not establish the required control for a nonprofit grant portal to detect policy violations emerging after deployment. The change is applied outside the component responsible for the requirement identified by a nonprofit grant portal. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
D:The suggested action—audit only before launch and ignore live behavior—does not establish the required control for a nonprofit grant portal to detect policy violations emerging after deployment. It replaces a verifiable engineering control with an untested shortcut in the a nonprofit grant portal deployment. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
Question 679
At an airline operations unit, engineers need to normalize inconsistent aliases before generating embeddings. The next release must be evaluated against the previous baseline. The initial assessment isolates this issue from unrelated infrastructure changes. Which remediation addresses the root problem?
- Store inconsistent aliases without preserving a canonical mapping
- Increase top-p to compensate for source inconsistency
- Remove all metadata because the input is inconsistent
- Apply deterministic text normalization and entity mapping in the preprocessing pipeline
Explanation
Correct Answer
D:Canonical preprocessing reduces semantically duplicate or mismatched indexed content. In an airline operations unit, the engineering team should verify this design against its stated objective to normalize inconsistent aliases before generating embeddings, including the requirement that the next release must be evaluated against the previous baseline.
Incorrect Answers
A:The suggested action—store inconsistent aliases without preserving a canonical mapping—does not establish the required control for an airline operations unit to normalize inconsistent aliases before generating embeddings. No demonstrated capability in this scenario guarantees that this step resolves the concern raised by an airline operations unit. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
B:The suggested action—increase top-p to compensate for source inconsistency—does not establish the required control for an airline operations unit to normalize inconsistent aliases before generating embeddings. It replaces a verifiable engineering control with an untested shortcut in the an airline operations unit deployment. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
C:The suggested action—remove all metadata because the input is inconsistent—does not establish the required control for an airline operations unit to normalize inconsistent aliases before generating embeddings. The change is applied outside the component responsible for the requirement identified by an airline operations unit. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
Question 680
At an automotive parts supplier, engineers need to assess agent workflows that must call external tools accurately. Operators need measurable acceptance criteria before rollout. A recent review found that the current design has no reliable safeguard for this case. What should the lead engineer recommend?
- Count total tool calls and assume more is better
- Check only the model's prose fluency
- Ignore side effects and consider any final answer a success
- Measure task success, tool-choice correctness, authorization failures and excessive steps
Explanation
Correct Answer
D:Agent evaluation must include execution behavior and real task outcomes. In an automotive parts supplier, the engineering team should verify this design against its stated objective to assess agent workflows that must call external tools accurately, including the requirement that operators need measurable acceptance criteria before rollout.
Incorrect Answers
A:The suggested action—count total tool calls and assume more is better—does not establish the required control for an automotive parts supplier to assess agent workflows that must call external tools accurately. No demonstrated capability in this scenario guarantees that this step resolves the concern raised by an automotive parts supplier. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
B:The suggested action—check only the model's prose fluency—does not establish the required control for an automotive parts supplier to assess agent workflows that must call external tools accurately. It replaces a verifiable engineering control with an untested shortcut in the an automotive parts supplier deployment. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.
C:The suggested action—ignore side effects and consider any final answer a success—does not establish the required control for an automotive parts supplier to assess agent workflows that must call external tools accurately. The change is applied outside the component responsible for the requirement identified by an automotive parts supplier. In this deployment, engineers should verify the missing mechanism rather than assume the proposed step satisfies the acceptance conditions.