Free Claude Certified Architect - Professional Practice Questions

A.  

Opus, because every moderation decision requires maximum reasoning depth regardless of classification complexity.

B.  

Haiku, because its latency and cost profile align with high-volume classification workloads that require limited reasoning depth.

C.  

Sonnet, because larger general-purpose models are preferred even when workload latency requirements are strict.

D.  

Sonnet with extended thinking enabled, because deeper reasoning should be applied to every moderation request to improve edge-case handling.

A.  

The most recent set of design presentation slides without component-level diagrams, interface contracts, an evaluation framework with a reference set, runbooks, or a known-limitations register.

B.  

A verbal walkthrough conducted on the day of handoff with no written architecture overview, ADRs, component contracts, evaluation framework, runbooks, playbook, or known limitations.

C.  

Architecture overview, ADRs, component contracts, evaluation framework with reference set, runbooks, on-call playbook, and known limitations.

D.  

Source code alone with no integrating architecture overview, ADRs, component contracts, evaluation framework, runbooks, on-call playbook, or known-limitations register to support the delivery team.

A.  

Approval rate parity does not by itself assess error rate parity across protected groups.

B.  

Match with human underwriters does not establish freedom from underwriter-introduced bias.

C.  

The 94 percent match rate is sufficient evidence of fairness for the assistant’s decisions.

D.  

The fairness team’s review process likely missed at least some of the protected groups studied.

E.  

A larger sample is needed before any meaningful fairness claim can be made about the model.

A.  

Trusting that the model will intrinsically recognize and reject all bypass attempts without prompt-level instructions, runtime classifiers, scoped permissions, or audit logging.

B.  

Prompt-level instructions that treat user content as untrusted data, runtime classifiers that detect override attempts, scoped tool permissions that cannot be elevated by user content, and audit logging of attempts.

C.  

Removing all safety restrictions and guardrails to eliminate the attack surface that bypass attempts target, accepting that this makes the assistant unrestricted for all inputs.

D.  

Granting users any privilege level they assert in their message content, on the assumption that cooperative behavior requires honoring self-declared authority without independent verification.

A.  

Adversarial tests that probe for prompt-injection vulnerabilities.

B.  

Regression tests scored against a stable reference set of known-good behavior.

C.  

Smoke tests that confirm high-level system availability after the change.

D.  

Integration tests that validate cross-component pipeline behavior.

A.  

Apply progressive tool discovery so a curated initial subset is exposed and additional tools are loaded on demand based on the task.

B.  

Use a separate model call to summarize all 60 tool definitions before each user turn.

C.  

Increase the maximum context length and load all 60 tool definitions on every request, accepting the higher token cost and latency as necessary for full capability.

D.  

Hard-code a fixed set of five tools per request to reduce token usage, regardless of whether those tools are relevant to the current task.

A.  

Determine the minimum detectable effect size and the sample size needed for power.

B.  

Decide whether to promote, reject, or iterate the candidate based on the analysis.

C.  

Define the hypothesis and the primary success metric for the comparison.

D.  

Analyze the results against the predefined success metric and significance threshold.

E.  

Document the recommendation, the trade-offs accepted, and the alternatives considered.

A.  

A runbook step requires a clarification edit to improve on-c all guidance accuracy, which can be handled as a documentation update without changes to component responsibilities or core contracts.

B.  

A dashboard alert threshold needs a small numerical adjustment to reduce false-positive noise, which can be handled as an operational configuration change without a new design cycle.

C.  

A minor copy edit is requested in a customer-facing string within the existing UI, which can be handled as a localized content change without altering component responsibilities or contracts.

D.  

The system’s current architecture cannot meet the new requirements without changes to component responsibilities or core contracts.