Amazon AWS Certified Generative AI Developer - Professional AIP-C01 Practice Test 40

Question 390

A team assumes prompt caching always hits. Which resource-efficiency decision fits the workload?

Inspect actual cache usage metrics and model/API limits, not prompt repetition alone.
Equate two semantically similar prompts with a guaranteed explicit prompt-cache hit.
Assume a cached hit exists whenever the application logs a prompt repetition.
Use generation throughput improvement alone as a cache hit metric.

Explanation

Correct Answer

Measure cache-write/cache-read token counters and actual hit rate with supported model/API limits before estimating cost or latency savings.

Incorrect Answers

Checkpoint caching requires matching eligible token prefixes.

Model, minimum tokens and TTL can cause misses.

Faster output can have other causes and does not prove cache token reuse.

Question 391

A finance review discovers that thousands of nightly summarization jobs use real-time synchronous Bedrock calls even though the output is needed only the next morning. What cost-and-deadline comparison should the team perform?

Submit a new synchronous streaming chat request for every night-time record despite a flexible batch deadline.
Assume batch jobs support live multi-turn tool calls in each record.
Compare eligible Bedrock asynchronous batch inference pricing, model/Region support and completion window against the current synchronous request spend and its unnecessary low-latency requirement.
Send the entire job as one interactive prompt and hope the model handles all records within context.

Explanation

Correct Answer

For noninteractive nightly workloads, compare supported S3-input/output batch jobs against synchronous invoke economics and actual delivery deadlines. Batch is not a per-turn replacement and has model, operation and Region eligibility limits.

Incorrect Answers

Interactive request overhead may be unnecessary at large scale.

Bedrock batch does not support interactive tool calling.

A single bounded-context dialogue is not a robust bulk processing pipeline.

Question 392

A small classification task routes to most expensive model. Which optimization balances actual cost and accepted task quality?

Use the smallest model for all requests because it has the lowest token price.
Switch based solely on model parameter count rather than representative labeled task tests.
Introduce quality-evaluated model tiering with measured cost per correct answer.
Optimize input tokens without measuring missing/incorrect output classifications.

Explanation

Correct Answer

Route classification to a cheaper evaluated model only when its measured class-level accuracy meets accepted thresholds and cost per correct answer improves.

Incorrect Answers

Incorrect answers may cost more than token savings.

Size does not reliably rank business classification performance.

Low inference usage alone is not the business objective.

Question 393

A complicated legal reasoning case routes to a tiny classifier. Which resource-efficiency decision fits the workload?

Always use a tiny classifier for legal exceptions because its per-call price is lower.
Label the output 'expert' to compensate for the model's missing task capability.
Drop the governing legal clauses to fit the small model's input limit.
Escalate complexity to a validated capable model rather than optimizing price alone.

Explanation

Correct Answer

Escalate complex legal reasoning to a capable, tested model and include required source/accuracy constraints rather than imposing a cheap-classifier route.

Incorrect Answers

A classifier may lack reasoning and evidence capacity for complex exceptions.

Marketing labels do not improve technical correctness.

Context truncation can remove decisive policy conditions.

Question 394

A prompt compression update drops an essential compliance clause. What should be measured before changing inference spending?

Adopt the lowest-token prompt automatically without running source-grounded acceptance cases.
Compare answer quality on held-out cases before adopting shorter context.
Compare compressed and full prompts only on a generic casual-chat benchmark.
Keep only the opening paragraph of long policies and assume exceptions never matter.

Explanation

Correct Answer

Evaluate the shortened prompt against held-out compliance clauses and reject compression that loses required constraints, even if tokens fall.

Incorrect Answers

A shorter context may omit legally material information.

Unrelated examples do not test compliance-clause retention.

Missing later clauses can reverse an answer's meaning.

Question 395

An application caches answers after an inventory price change. Which optimization balances actual cost and accepted task quality?

Keep an expired cached price indefinitely if its text is similar to the new inventory query.
Change only the FM temperature and keep returning old cached prices.
Trust the cached display price as final even after backend stock/price changes.
Set source-aware invalidation/TTL and require fresh price at transaction time.

Explanation

Correct Answer

Invalidate price caches on inventory updates and verify current price at purchase time; stale reuse can cause incorrect financial actions.

Incorrect Answers

Semantic similarity does not establish current price or stock.

Sampling cannot refresh a stale source of record.

A transaction needs authoritative current business values.

Question 396

A semantic cache returns one user's private result to another. Which resource-efficiency decision fits the workload?

Use one global semantic cache across all customers with only a topic-based key.
Hide names in cached text while leaving confidential account amounts and record IDs.
Scope cache keys and data permissions to identity and authorization context.
Rely on the model to recognize which cached answer belongs to the signed-in customer.

Explanation

Correct Answer

Partition semantic cache by authenticated user, tenant and effective data privileges, and prevent cross-actor retrieval of private responses.

Incorrect Answers

Similar questions can contain different private results.

Partial redaction may still disclose protected data.

A model classification is not server-side cache authorization.

Question 397

A request identifier cache treats paraphrases as identical transactions. What should be measured before changing inference spending?

Consider paraphrased payment requests identical and reuse an earlier payment receipt.
Separate semantic answer reuse from idempotent business-operation identity.
Generate new charge IDs on every retry because the new phrasing is slightly different.
Mark a business write complete when a similar cached natural-language reply is found.

Explanation

Correct Answer

Use semantic cache only for safe read-only answers; transactional idempotency must be tied to a stable operation ID and downstream commit state.

Incorrect Answers

Semantically similar words can still describe separate transactions.

Changing identity defeats idempotency and may duplicate charges.

Answer caches cannot prove the current operation was committed.

Question 398

An existing Bedrock deployment has been paying for committed throughput while production utilization remains far below capacity for several billing cycles. What optimization assessment should be made before altering the purchased capacity?

Keep paying for unused reserved throughput because its maximum capacity is high.
Switch blindly to on-demand when the model lacks supported on-demand inference in the target Region.
Measure only HTTP call count and ignore the token throughput that the reservation covers.
Compare paid commitment and observed reserved-throughput utilization with model- and Region-supported on-demand options, validating cost, traffic and service-level impacts.

Explanation

Correct Answer

Use measured low utilization and the current commitment to model reserved-capacity spend versus actually supported alternatives. A switch based only on a model name may fail if on-demand inference is unavailable in that Region.

Incorrect Answers

Low utilization may make committed spending wasteful.

Deployment-mode availability constrains valid alternatives.

Capacity and pricing depend on actual token workload, not request totals alone.

Question 399

Token usage spikes after a prompt deployment. Which resource-efficiency decision fits the workload?

Attribute all new token spend to user growth without measuring per-request token lengths.
Attribute input/output token deltas to prompt version and request class.
Look only at total billing and ignore whether extra tokens are input or output.
Remove version identifiers from telemetry to reduce log volume.

Explanation

Correct Answer

Segment input/output token counters by prompt version and request class to isolate which deployment change increased spend.

Incorrect Answers

Prompt regression may increase token cost independently of traffic volume.

Different token categories can have different rates and causes.

Without version tags, the spike cannot be tied to the prompt release.

Question 400

An app pays to generate identical static onboarding text repeatedly. What should be measured before changing inference spending?

Cache personalized account status for every user under one shared onboarding key.
Precompute approved immutable responses while retaining dynamic-policy checks.
Regenerate identical permanent boilerplate on every request despite no variable input.
Present precomputed content forever without revalidation when source policy changes.

Explanation

Correct Answer

Precompute approved immutable onboarding copy, while validating current policy exceptions and avoiding stale dynamic state in reused responses.

Incorrect Answers

Static template reuse must not leak private or mutable account data.

Unnecessary model calls inflate cost and latency.

A static answer must be invalidated when its policy basis changes.

Pass Your Amazon AWS Certified Generative AI Developer - Professional AIP-C01 Exam Easy!

Question 390

Explanation

Correct Answer

Incorrect Answers

Question 391

Explanation

Correct Answer

Incorrect Answers

Question 392

Explanation

Correct Answer

Incorrect Answers

Question 393

Explanation

Correct Answer

Incorrect Answers

Question 394

Explanation

Correct Answer

Incorrect Answers

Question 395

Explanation

Correct Answer

Incorrect Answers

Question 396

Explanation

Correct Answer

Incorrect Answers

Question 397

Explanation

Correct Answer

Incorrect Answers

Question 398

Explanation

Correct Answer

Incorrect Answers

Question 399

Explanation

Correct Answer

Incorrect Answers

Question 400

Explanation

Correct Answer

Incorrect Answers

AWS Certified Generative AI Developer - Professional AIP-C01 Practice Tests