Amazon AWS AIP-C01 GenAI Performance Optimization Practice Test

 

Topic 17 focuses on GenAI Application Performance Optimization for the AWS Certified Generative AI Developer – Professional certification and the AIP-C01 exam, using Amazon Bedrock and AWS generative AI services and architecture scenarios where relevant. For broader exam preparation, review the AWS Certified Generative AI Developer – Professional AIP-C01 Exam Dumps page. Each question includes a concise explanation of the correct answer and the technical reason the other choices are incorrect.

Question 1

The procurement automation group has a design goal to serve frequently requested deterministic results without invoking a model every time. What should the team choose?

1. concurrent invocation management
2. pre-computation for predictable queries
3. latency-optimized Bedrock inference
4. performance benchmarking

Correct Answer: B

 

Correct Answer

Answer B is correct because pre-computation for predictable queries is designed to serve frequently requested deterministic results without invoking a model every time. It computes or refreshes known outputs ahead of demand so user-facing latency is reduced.

Incorrect Answers

Answer A is incorrect because concurrent invocation management is primarily used to control the number of simultaneous model requests so throughput rises without overwhelming quotas or downstream services, which is a different requirement from the one being tested.

Answer C is incorrect because latency-optimized Bedrock inference is primarily used to reduce model response latency for an interactive workload when supported, which is a different requirement from the one being tested.

Answer D is incorrect because performance benchmarking is primarily used to compare latency, throughput, and quality before choosing an optimization, which is a different requirement from the one being tested.

 

Question 2

The observability engineering group needs an architecture that can reduce model response latency for an interactive workload when supported. Which choice best meets that need?

1. parallel model requests
2. temperature and top-p/top-k tuning
3. latency-optimized Bedrock inference
4. vector index optimization

Correct Answer: C

 

Correct Answer

Answer C is correct because latency-optimized Bedrock inference is designed to reduce model response latency for an interactive workload when supported. It selects latency-optimized model or inference capabilities and validates the quality-cost tradeoff for the use case.

Incorrect Answers

Answer A is incorrect because parallel model requests is primarily used to shorten a workflow containing independent model calls, which is a different requirement from the one being tested.

Answer B is incorrect because temperature and top-p/top-k tuning is primarily used to adjust generation diversity and determinism for the response-quality objective, which is a different requirement from the one being tested.

Answer D is incorrect because vector index optimization is primarily used to reduce retrieval latency for a large semantic-search corpus, which is a different requirement from the one being tested.

 

Question 3

The managed services operations team needs an architecture that can shorten a workflow containing independent model calls. Which choice best meets that need?

1. token-capacity planning
2. parallel model requests
3. query preprocessing
4. response streaming

Correct Answer: B

 

Correct Answer

Answer B is correct because parallel model requests is designed to shorten a workflow containing independent model calls. It runs independent inference steps concurrently instead of serializing work that has no dependency.

Incorrect Answers

Answer A is incorrect because token-capacity planning is primarily used to size infrastructure and quotas around prompt and completion processing demand, which is a different requirement from the one being tested.

Answer C is incorrect because query preprocessing is primarily used to improve retrieval efficiency by normalizing or simplifying search requests before execution, which is a different requirement from the one being tested.

Answer D is incorrect because response streaming is primarily used to improve perceived latency by showing generated output before inference is fully complete, which is a different requirement from the one being tested.

 

Question 4

The service management platform must improve perceived latency by showing generated output before inference is fully complete. Which approach is the strongest fit?

1. hybrid search with custom scoring
2. performance benchmarking
3. FM API call profiling
4. response streaming

Correct Answer: D

 

Correct Answer

Answer D is correct because response streaming is designed to improve perceived latency by showing generated output before inference is fully complete. It delivers incremental chunks or tokens to the client while generation continues.

Incorrect Answers

Answer A is incorrect because hybrid search with custom scoring is primarily used to improve retrieval relevance when both exact terms and semantic similarity matter, which is a different requirement from the one being tested.

Answer B is incorrect because performance benchmarking is primarily used to compare latency, throughput, and quality before choosing an optimization, which is a different requirement from the one being tested.

Answer C is incorrect because FM API call profiling is primarily used to identify latency concentrated in model calls, retries, serialization, or surrounding services, which is a different requirement from the one being tested.

 

Question 5

The digital experience platform needs an architecture that can compare latency, throughput, and quality before choosing an optimization. Which choice best meets that need?

1. performance benchmarking
2. vector database query optimization
3. vector index optimization
4. batch inference

Correct Answer: A

 

Correct Answer

Answer A is correct because performance benchmarking is designed to compare latency, throughput, and quality before choosing an optimization. It uses repeatable representative workloads so design changes can be evaluated quantitatively.

Incorrect Answers

Answer B is incorrect because vector database query optimization is primarily used to reduce retrieval time and wasted search work in a RAG pipeline, which is a different requirement from the one being tested.

Answer C is incorrect because vector index optimization is primarily used to reduce retrieval latency for a large semantic-search corpus, which is a different requirement from the one being tested.

Answer D is incorrect because batch inference is primarily used to process non-interactive inference workloads efficiently at scale, which is a different requirement from the one being tested.

 

Question 6

The cloud governance office has a design goal to reduce retrieval latency for a large semantic-search corpus. What should the team choose?

1. query preprocessing
2. vector index optimization
3. pre-computation for predictable queries
4. concurrent invocation management

Correct Answer: B

 

Correct Answer

Answer B is correct because vector index optimization is designed to reduce retrieval latency for a large semantic-search corpus. It tunes index type, dimensions, partitions, shards, or other search parameters based on measured query patterns.

Incorrect Answers

Answer A is incorrect because query preprocessing is primarily used to improve retrieval efficiency by normalizing or simplifying search requests before execution, which is a different requirement from the one being tested.

Answer C is incorrect because pre-computation for predictable queries is primarily used to serve frequently requested deterministic results without invoking a model every time, which is a different requirement from the one being tested.

Answer D is incorrect because concurrent invocation management is primarily used to control the number of simultaneous model requests so throughput rises without overwhelming quotas or downstream services, which is a different requirement from the one being tested.

 

Question 7

The travel booking application has a design goal to improve retrieval efficiency by normalizing or simplifying search requests before execution. What should the team choose?

1. temperature and top-p/top-k tuning
2. latency-optimized Bedrock inference
3. hybrid search with custom scoring
4. query preprocessing

Correct Answer: D

 

Correct Answer

Answer D is correct because query preprocessing is designed to improve retrieval efficiency by normalizing or simplifying search requests before execution. It cleans, reformulates, or classifies queries so the retrieval system performs less irrelevant work.

Incorrect Answers

Answer A is incorrect because temperature and top-p/top-k tuning is primarily used to adjust generation diversity and determinism for the response-quality objective, which is a different requirement from the one being tested.

Answer B is incorrect because latency-optimized Bedrock inference is primarily used to reduce model response latency for an interactive workload when supported, which is a different requirement from the one being tested.

Answer C is incorrect because hybrid search with custom scoring is primarily used to improve retrieval relevance when both exact terms and semantic similarity matter, which is a different requirement from the one being tested.

 

Question 8

The compliance engineering group is prioritizing a requirement to improve retrieval relevance when both exact terms and semantic similarity matter. Which implementation is most appropriate?

1. batch inference
2. token-capacity planning
3. hybrid search with custom scoring
4. parallel model requests

Correct Answer: C

 

Correct Answer

Answer C is correct because hybrid search with custom scoring is designed to improve retrieval relevance when both exact terms and semantic similarity matter. It combines lexical and vector signals with weighting or reranking appropriate to the domain.

Incorrect Answers

Answer A is incorrect because batch inference is primarily used to process non-interactive inference workloads efficiently at scale, which is a different requirement from the one being tested.

Answer B is incorrect because token-capacity planning is primarily used to size infrastructure and quotas around prompt and completion processing demand, which is a different requirement from the one being tested.

Answer D is incorrect because parallel model requests is primarily used to shorten a workflow containing independent model calls, which is a different requirement from the one being tested.

 

Question 9

The document intelligence team needs an architecture that can process non-interactive inference workloads efficiently at scale. Which choice best meets that need?

1. batch inference
2. FM API call profiling
3. concurrent invocation management
4. response streaming

Correct Answer: A

 

Correct Answer

Answer A is correct because batch inference is designed to process non-interactive inference workloads efficiently at scale. It groups queued inputs for throughput-oriented processing when immediate per-request responses are not required.

Incorrect Answers

Answer B is incorrect because FM API call profiling is primarily used to identify latency concentrated in model calls, retries, serialization, or surrounding services, which is a different requirement from the one being tested.

Answer C is incorrect because concurrent invocation management is primarily used to control the number of simultaneous model requests so throughput rises without overwhelming quotas or downstream services, which is a different requirement from the one being tested.

Answer D is incorrect because response streaming is primarily used to improve perceived latency by showing generated output before inference is fully complete, which is a different requirement from the one being tested.

 

Question 10

The human resources automation team needs an architecture that can control the number of simultaneous model requests so throughput rises without overwhelming quotas or downstream services. Which choice best meets that need?

1. vector database query optimization
2. temperature and top-p/top-k tuning
3. concurrent invocation management
4. performance benchmarking

Correct Answer: C

 

Correct Answer

Answer C is correct because concurrent invocation management is designed to control the number of simultaneous model requests so throughput rises without overwhelming quotas or downstream services. It uses bounded concurrency and backpressure rather than allowing unregulated parallel calls.

Incorrect Answers

Answer A is incorrect because vector database query optimization is primarily used to reduce retrieval time and wasted search work in a RAG pipeline, which is a different requirement from the one being tested.

Answer B is incorrect because temperature and top-p/top-k tuning is primarily used to adjust generation diversity and determinism for the response-quality objective, which is a different requirement from the one being tested.

Answer D is incorrect because performance benchmarking is primarily used to compare latency, throughput, and quality before choosing an optimization, which is a different requirement from the one being tested.

 

Question 11

The global infrastructure team has a design goal to adjust generation diversity and determinism for the response-quality objective. What should the team choose?

1. temperature and top-p/top-k tuning
2. vector index optimization
3. token-capacity planning
4. pre-computation for predictable queries

Correct Answer: A

 

Correct Answer

Answer A is correct because temperature and top-p/top-k tuning is designed to adjust generation diversity and determinism for the response-quality objective. It changes model sampling behavior and validates the effect with representative evaluations.

Incorrect Answers

Answer B is incorrect because vector index optimization is primarily used to reduce retrieval latency for a large semantic-search corpus, which is a different requirement from the one being tested.

Answer C is incorrect because token-capacity planning is primarily used to size infrastructure and quotas around prompt and completion processing demand, which is a different requirement from the one being tested.

Answer D is incorrect because pre-computation for predictable queries is primarily used to serve frequently requested deterministic results without invoking a model every time, which is a different requirement from the one being tested.

 

Question 12

The legal compliance automation group has a design goal to size infrastructure and quotas around prompt and completion processing demand. What should the team choose?

1. latency-optimized Bedrock inference
2. query preprocessing
3. token-capacity planning
4. FM API call profiling

Correct Answer: C

 

Correct Answer

Answer C is correct because token-capacity planning is designed to size infrastructure and quotas around prompt and completion processing demand. It models expected token rates and concurrency instead of considering only http request counts.

Incorrect Answers

Answer A is incorrect because latency-optimized Bedrock inference is primarily used to reduce model response latency for an interactive workload when supported, which is a different requirement from the one being tested.

Answer B is incorrect because query preprocessing is primarily used to improve retrieval efficiency by normalizing or simplifying search requests before execution, which is a different requirement from the one being tested.

Answer D is incorrect because FM API call profiling is primarily used to identify latency concentrated in model calls, retries, serialization, or surrounding services, which is a different requirement from the one being tested.

 

Question 13

The developer experience platform needs an architecture that can identify latency concentrated in model calls, retries, serialization, or surrounding services. Which choice best meets that need?

1. vector database query optimization
2. FM API call profiling
3. parallel model requests
4. hybrid search with custom scoring

Correct Answer: B

 

Correct Answer

Answer B is correct because FM API call profiling is designed to identify latency concentrated in model calls, retries, serialization, or surrounding services. It measures the end-to-end request path so optimization targets the actual bottleneck.

Incorrect Answers

Answer A is incorrect because vector database query optimization is primarily used to reduce retrieval time and wasted search work in a RAG pipeline, which is a different requirement from the one being tested.

Answer C is incorrect because parallel model requests is primarily used to shorten a workflow containing independent model calls, which is a different requirement from the one being tested.

Answer D is incorrect because hybrid search with custom scoring is primarily used to improve retrieval relevance when both exact terms and semantic similarity matter, which is a different requirement from the one being tested.

 

Question 14

The global support organization must reduce retrieval time and wasted search work in a RAG pipeline. Which approach is the strongest fit?

1. vector database query optimization
2. batch inference
3. pre-computation for predictable queries
4. response streaming

Correct Answer: A

 

Correct Answer

Answer A is correct because vector database query optimization is designed to reduce retrieval time and wasted search work in a RAG pipeline. It uses query plans, filters, index tuning, and result limits suited to the semantic search workload.

Incorrect Answers

Answer B is incorrect because batch inference is primarily used to process non-interactive inference workloads efficiently at scale, which is a different requirement from the one being tested.

Answer C is incorrect because pre-computation for predictable queries is primarily used to serve frequently requested deterministic results without invoking a model every time, which is a different requirement from the one being tested.

Answer D is incorrect because response streaming is primarily used to improve perceived latency by showing generated output before inference is fully complete, which is a different requirement from the one being tested.

 

Question 15

The digital experience platform needs to identify a capability with this behavior: computes or refreshes known outputs ahead of demand so user-facing latency is reduced. What is the best match?

1. performance benchmarking
2. pre-computation for predictable queries
3. concurrent invocation management
4. latency-optimized Bedrock inference

Correct Answer: B

 

Correct Answer

Answer B is correct because the description directly matches pre-computation for predictable queries. pre-computation for predictable queries computes or refreshes known outputs ahead of demand so user-facing latency is reduced.

Incorrect Answers

Answer A is incorrect because performance benchmarking is primarily used to compare latency, throughput, and quality before choosing an optimization, which is a different requirement from the one being tested.

Answer C is incorrect because concurrent invocation management is primarily used to control the number of simultaneous model requests so throughput rises without overwhelming quotas or downstream services, which is a different requirement from the one being tested.

Answer D is incorrect because latency-optimized Bedrock inference is primarily used to reduce model response latency for an interactive workload when supported, which is a different requirement from the one being tested.

 

Question 16

The cloud governance office needs to identify a capability with this behavior: selects latency-optimized model or inference capabilities and validates the quality-cost tradeoff for the use case. What is the best match?

1. vector index optimization
2. parallel model requests
3. latency-optimized Bedrock inference
4. temperature and top-p/top-k tuning

Correct Answer: C

 

Correct Answer

Answer C is correct because the description directly matches latency-optimized Bedrock inference. latency-optimized Bedrock inference selects latency-optimized model or inference capabilities and validates the quality-cost tradeoff for the use case.

Incorrect Answers

Answer A is incorrect because vector index optimization is primarily used to reduce retrieval latency for a large semantic-search corpus, which is a different requirement from the one being tested.

Answer B is incorrect because parallel model requests is primarily used to shorten a workflow containing independent model calls, which is a different requirement from the one being tested.

Answer D is incorrect because temperature and top-p/top-k tuning is primarily used to adjust generation diversity and determinism for the response-quality objective, which is a different requirement from the one being tested.

 

Question 17

The travel booking application describes a component that runs independent inference steps concurrently instead of serializing work that has no dependency. Which capability is being described?

1. token-capacity planning
2. parallel model requests
3. response streaming
4. query preprocessing

Correct Answer: B

 

Correct Answer

Answer B is correct because the description directly matches parallel model requests. parallel model requests runs independent inference steps concurrently instead of serializing work that has no dependency.

Incorrect Answers

Answer A is incorrect because token-capacity planning is primarily used to size infrastructure and quotas around prompt and completion processing demand, which is a different requirement from the one being tested.

Answer C is incorrect because response streaming is primarily used to improve perceived latency by showing generated output before inference is fully complete, which is a different requirement from the one being tested.

Answer D is incorrect because query preprocessing is primarily used to improve retrieval efficiency by normalizing or simplifying search requests before execution, which is a different requirement from the one being tested.

 

Question 18

The compliance engineering group documents this GenAI behavior: delivers incremental chunks or tokens to the client while generation continues. Which capability matches it?

1. response streaming
2. performance benchmarking
3. FM API call profiling
4. hybrid search with custom scoring

Correct Answer: A

 

Correct Answer

Answer A is correct because the description directly matches response streaming. response streaming delivers incremental chunks or tokens to the client while generation continues.

Incorrect Answers

Answer B is incorrect because performance benchmarking is primarily used to compare latency, throughput, and quality before choosing an optimization, which is a different requirement from the one being tested.

Answer C is incorrect because FM API call profiling is primarily used to identify latency concentrated in model calls, retries, serialization, or surrounding services, which is a different requirement from the one being tested.

Answer D is incorrect because hybrid search with custom scoring is primarily used to improve retrieval relevance when both exact terms and semantic similarity matter, which is a different requirement from the one being tested.

 

Question 19

The document intelligence team documents this GenAI behavior: uses repeatable representative workloads so design changes can be evaluated quantitatively. Which capability matches it?

1. vector index optimization
2. vector database query optimization
3. performance benchmarking
4. batch inference

Correct Answer: C

 

Correct Answer

Answer C is correct because the description directly matches performance benchmarking. performance benchmarking uses repeatable representative workloads so design changes can be evaluated quantitatively.

Incorrect Answers

Answer A is incorrect because vector index optimization is primarily used to reduce retrieval latency for a large semantic-search corpus, which is a different requirement from the one being tested.

Answer B is incorrect because vector database query optimization is primarily used to reduce retrieval time and wasted search work in a RAG pipeline, which is a different requirement from the one being tested.

Answer D is incorrect because batch inference is primarily used to process non-interactive inference workloads efficiently at scale, which is a different requirement from the one being tested.

 

Question 20

The human resources automation team needs to identify a capability with this behavior: tunes index type, dimensions, partitions, shards, or other search parameters based on measured query patterns. What is the best match?

1. vector index optimization
2. query preprocessing
3. concurrent invocation management
4. pre-computation for predictable queries

Correct Answer: A

 

Correct Answer

Answer A is correct because the description directly matches vector index optimization. vector index optimization tunes index type, dimensions, partitions, shards, or other search parameters based on measured query patterns.

Incorrect Answers

Answer B is incorrect because query preprocessing is primarily used to improve retrieval efficiency by normalizing or simplifying search requests before execution, which is a different requirement from the one being tested.

Answer C is incorrect because concurrent invocation management is primarily used to control the number of simultaneous model requests so throughput rises without overwhelming quotas or downstream services, which is a different requirement from the one being tested.

Answer D is incorrect because pre-computation for predictable queries is primarily used to serve frequently requested deterministic results without invoking a model every time, which is a different requirement from the one being tested.

 

Question 21

Within the global infrastructure team’s architecture, which capability matches this technical description: cleans, reformulates, or classifies queries so the retrieval system performs less irrelevant work?

1. hybrid search with custom scoring
2. latency-optimized Bedrock inference
3. query preprocessing
4. temperature and top-p/top-k tuning

Correct Answer: C

 

Correct Answer

Answer C is correct because the description directly matches query preprocessing. query preprocessing cleans, reformulates, or classifies queries so the retrieval system performs less irrelevant work.

Incorrect Answers

Answer A is incorrect because hybrid search with custom scoring is primarily used to improve retrieval relevance when both exact terms and semantic similarity matter, which is a different requirement from the one being tested.

Answer B is incorrect because latency-optimized Bedrock inference is primarily used to reduce model response latency for an interactive workload when supported, which is a different requirement from the one being tested.

Answer D is incorrect because temperature and top-p/top-k tuning is primarily used to adjust generation diversity and determinism for the response-quality objective, which is a different requirement from the one being tested.

 

Question 22

The legal compliance automation group needs to identify a capability with this behavior: combines lexical and vector signals with weighting or reranking appropriate to the domain. What is the best match?

1. batch inference
2. hybrid search with custom scoring
3. parallel model requests
4. token-capacity planning

Correct Answer: B

 

Correct Answer

Answer B is correct because the description directly matches hybrid search with custom scoring. hybrid search with custom scoring combines lexical and vector signals with weighting or reranking appropriate to the domain.

Incorrect Answers

Answer A is incorrect because batch inference is primarily used to process non-interactive inference workloads efficiently at scale, which is a different requirement from the one being tested.

Answer C is incorrect because parallel model requests is primarily used to shorten a workflow containing independent model calls, which is a different requirement from the one being tested.

Answer D is incorrect because token-capacity planning is primarily used to size infrastructure and quotas around prompt and completion processing demand, which is a different requirement from the one being tested.

 

Question 23

The developer experience platform documents this GenAI behavior: groups queued inputs for throughput-oriented processing when immediate per-request responses are not required. Which capability matches it?

1. response streaming
2. FM API call profiling
3. concurrent invocation management
4. batch inference

Correct Answer: D

 

Correct Answer

Answer D is correct because the description directly matches batch inference. batch inference groups queued inputs for throughput-oriented processing when immediate per-request responses are not required.

Incorrect Answers

Answer A is incorrect because response streaming is primarily used to improve perceived latency by showing generated output before inference is fully complete, which is a different requirement from the one being tested.

Answer B is incorrect because FM API call profiling is primarily used to identify latency concentrated in model calls, retries, serialization, or surrounding services, which is a different requirement from the one being tested.

Answer C is incorrect because concurrent invocation management is primarily used to control the number of simultaneous model requests so throughput rises without overwhelming quotas or downstream services, which is a different requirement from the one being tested.

 

Question 24

The global support organization describes a component that uses bounded concurrency and backpressure rather than allowing unregulated parallel calls. Which capability is being described?

1. temperature and top-p/top-k tuning
2. vector database query optimization
3. performance benchmarking
4. concurrent invocation management

Correct Answer: D

 

Correct Answer

Answer D is correct because the description directly matches concurrent invocation management. concurrent invocation management uses bounded concurrency and backpressure rather than allowing unregulated parallel calls.

Incorrect Answers

Answer A is incorrect because temperature and top-p/top-k tuning is primarily used to adjust generation diversity and determinism for the response-quality objective, which is a different requirement from the one being tested.

Answer B is incorrect because vector database query optimization is primarily used to reduce retrieval time and wasted search work in a RAG pipeline, which is a different requirement from the one being tested.

Answer C is incorrect because performance benchmarking is primarily used to compare latency, throughput, and quality before choosing an optimization, which is a different requirement from the one being tested.

 

Question 25

Within the data governance group’s architecture, which capability matches this technical description: changes model sampling behavior and validates the effect with representative evaluations?

1. token-capacity planning
2. temperature and top-p/top-k tuning
3. vector index optimization
4. pre-computation for predictable queries

Correct Answer: B

 

Correct Answer

Answer B is correct because the description directly matches temperature and top-p/top-k tuning. temperature and top-p/top-k tuning changes model sampling behavior and validates the effect with representative evaluations.

Incorrect Answers

Answer A is incorrect because token-capacity planning is primarily used to size infrastructure and quotas around prompt and completion processing demand, which is a different requirement from the one being tested.

Answer C is incorrect because vector index optimization is primarily used to reduce retrieval latency for a large semantic-search corpus, which is a different requirement from the one being tested.

Answer D is incorrect because pre-computation for predictable queries is primarily used to serve frequently requested deterministic results without invoking a model every time, which is a different requirement from the one being tested.

 

Question 26

Within the enterprise architecture board’s architecture, which capability matches this technical description: models expected token rates and concurrency instead of considering only HTTP request counts?

1. latency-optimized Bedrock inference
2. FM API call profiling
3. token-capacity planning
4. query preprocessing

Correct Answer: C

 

Correct Answer

Answer C is correct because the description directly matches token-capacity planning. token-capacity planning models expected token rates and concurrency instead of considering only HTTP request counts.

Incorrect Answers

Answer A is incorrect because latency-optimized Bedrock inference is primarily used to reduce model response latency for an interactive workload when supported, which is a different requirement from the one being tested.

Answer B is incorrect because FM API call profiling is primarily used to identify latency concentrated in model calls, retries, serialization, or surrounding services, which is a different requirement from the one being tested.

Answer D is incorrect because query preprocessing is primarily used to improve retrieval efficiency by normalizing or simplifying search requests before execution, which is a different requirement from the one being tested.

 

Question 27

The digital banking platform documents this GenAI behavior: measures the end-to-end request path so optimization targets the actual bottleneck. Which capability matches it?

1. parallel model requests
2. vector database query optimization
3. hybrid search with custom scoring
4. FM API call profiling

Correct Answer: D

 

Correct Answer

Answer D is correct because the description directly matches FM API call profiling. FM API call profiling measures the end-to-end request path so optimization targets the actual bottleneck.

Incorrect Answers

Answer A is incorrect because parallel model requests is primarily used to shorten a workflow containing independent model calls, which is a different requirement from the one being tested.

Answer B is incorrect because vector database query optimization is primarily used to reduce retrieval time and wasted search work in a RAG pipeline, which is a different requirement from the one being tested.

Answer C is incorrect because hybrid search with custom scoring is primarily used to improve retrieval relevance when both exact terms and semantic similarity matter, which is a different requirement from the one being tested.

 

Question 28

The data science enablement group describes a component that uses query plans, filters, index tuning, and result limits suited to the semantic search workload. Which capability is being described?

1. vector database query optimization
2. pre-computation for predictable queries
3. response streaming
4. batch inference

Correct Answer: A

 

Correct Answer

Answer A is correct because the description directly matches vector database query optimization. vector database query optimization uses query plans, filters, index tuning, and result limits suited to the semantic search workload.

Incorrect Answers

Answer B is incorrect because pre-computation for predictable queries is primarily used to serve frequently requested deterministic results without invoking a model every time, which is a different requirement from the one being tested.

Answer C is incorrect because response streaming is primarily used to improve perceived latency by showing generated output before inference is fully complete, which is a different requirement from the one being tested.

Answer D is incorrect because batch inference is primarily used to process non-interactive inference workloads efficiently at scale, which is a different requirement from the one being tested.

 

Question 29

The document intelligence team plans to adopt pre-computation for predictable queries. Which outcome should drive that decision?

1. Serve frequently requested deterministic results without invoking a model every time
2. Control the number of simultaneous model requests so throughput rises without overwhelming quotas or downstream services
3. Compare latency, throughput, and quality before choosing an optimization
4. Reduce model response latency for an interactive workload when supported

Correct Answer: A

 

Correct Answer

Answer A is correct because pre-computation for predictable queries is specifically used to serve frequently requested deterministic results without invoking a model every time. It computes or refreshes known outputs ahead of demand so user-facing latency is reduced.

Incorrect Answers

Answer B is incorrect because that requirement aligns with concurrent invocation management, not pre-computation for predictable queries.

Answer C is incorrect because that requirement aligns with performance benchmarking, not pre-computation for predictable queries.

Answer D is incorrect because that requirement aligns with latency-optimized Bedrock inference, not pre-computation for predictable queries.

 

Question 30

Within the human resources automation team’s design, the team highlights latency-optimized Bedrock inference. Which need does that component address?

1. Shorten a workflow containing independent model calls
2. Reduce retrieval latency for a large semantic-search corpus
3. Reduce model response latency for an interactive workload when supported
4. Adjust generation diversity and determinism for the response-quality objective

Correct Answer: C

 

Correct Answer

Answer C is correct because latency-optimized Bedrock inference is specifically used to reduce model response latency for an interactive workload when supported. It selects latency-optimized model or inference capabilities and validates the quality-cost tradeoff for the use case.

Incorrect Answers

Answer A is incorrect because that requirement aligns with parallel model requests, not latency-optimized Bedrock inference.

Answer B is incorrect because that requirement aligns with vector index optimization, not latency-optimized Bedrock inference.

Answer D is incorrect because that requirement aligns with temperature and top-p/top-k tuning, not latency-optimized Bedrock inference.

 

Question 31

Why would the global infrastructure team introduce parallel model requests into the GenAI architecture?

1. Size infrastructure and quotas around prompt and completion processing demand
2. Shorten a workflow containing independent model calls
3. Improve retrieval efficiency by normalizing or simplifying search requests before execution
4. Improve perceived latency by showing generated output before inference is fully complete

Correct Answer: B

 

Correct Answer

Answer B is correct because parallel model requests is specifically used to shorten a workflow containing independent model calls. It runs independent inference steps concurrently instead of serializing work that has no dependency.

Incorrect Answers

Answer A is incorrect because that requirement aligns with token-capacity planning, not parallel model requests.

Answer C is incorrect because that requirement aligns with query preprocessing, not parallel model requests.

Answer D is incorrect because that requirement aligns with response streaming, not parallel model requests.

 

Question 32

The legal compliance automation group has proposed response streaming for its design. Which requirement best justifies it?

1. Improve perceived latency by showing generated output before inference is fully complete
2. Compare latency, throughput, and quality before choosing an optimization
3. Identify latency concentrated in model calls, retries, serialization, or surrounding services
4. Improve retrieval relevance when both exact terms and semantic similarity matter

Correct Answer: A

 

Correct Answer

Answer A is correct because response streaming is specifically used to improve perceived latency by showing generated output before inference is fully complete. It delivers incremental chunks or tokens to the client while generation continues.

Incorrect Answers

Answer B is incorrect because that requirement aligns with performance benchmarking, not response streaming.

Answer C is incorrect because that requirement aligns with FM API call profiling, not response streaming.

Answer D is incorrect because that requirement aligns with hybrid search with custom scoring, not response streaming.

 

Question 33

The developer experience platform plans to adopt performance benchmarking. Which outcome should drive that decision?

1. Reduce retrieval latency for a large semantic-search corpus
2. Process non-interactive inference workloads efficiently at scale
3. Reduce retrieval time and wasted search work in a RAG pipeline
4. Compare latency, throughput, and quality before choosing an optimization

Correct Answer: D

 

Correct Answer

Answer D is correct because performance benchmarking is specifically used to compare latency, throughput, and quality before choosing an optimization. It uses repeatable representative workloads so design changes can be evaluated quantitatively.

Incorrect Answers

Answer A is incorrect because that requirement aligns with vector index optimization, not performance benchmarking.

Answer B is incorrect because that requirement aligns with batch inference, not performance benchmarking.

Answer C is incorrect because that requirement aligns with vector database query optimization, not performance benchmarking.

 

Question 34

The global support organization has proposed vector index optimization for its design. Which requirement best justifies it?

1. Improve retrieval efficiency by normalizing or simplifying search requests before execution
2. Control the number of simultaneous model requests so throughput rises without overwhelming quotas or downstream services
3. Serve frequently requested deterministic results without invoking a model every time
4. Reduce retrieval latency for a large semantic-search corpus

Correct Answer: D

 

Correct Answer

Answer D is correct because vector index optimization is specifically used to reduce retrieval latency for a large semantic-search corpus. It tunes index type, dimensions, partitions, shards, or other search parameters based on measured query patterns.

Incorrect Answers

Answer A is incorrect because that requirement aligns with query preprocessing, not vector index optimization.

Answer B is incorrect because that requirement aligns with concurrent invocation management, not vector index optimization.

Answer C is incorrect because that requirement aligns with pre-computation for predictable queries, not vector index optimization.

 

Question 35

The data governance group has proposed query preprocessing for its design. Which requirement best justifies it?

1. Adjust generation diversity and determinism for the response-quality objective
2. Improve retrieval relevance when both exact terms and semantic similarity matter
3. Reduce model response latency for an interactive workload when supported
4. Improve retrieval efficiency by normalizing or simplifying search requests before execution

Correct Answer: D

 

Correct Answer

Answer D is correct because query preprocessing is specifically used to improve retrieval efficiency by normalizing or simplifying search requests before execution. It cleans, reformulates, or classifies queries so the retrieval system performs less irrelevant work.

Incorrect Answers

Answer A is incorrect because that requirement aligns with temperature and top-p/top-k tuning, not query preprocessing.

Answer B is incorrect because that requirement aligns with hybrid search with custom scoring, not query preprocessing.

Answer C is incorrect because that requirement aligns with latency-optimized Bedrock inference, not query preprocessing.