Amazon AWS AIP-C01 Foundation Model Deployment Practice Test

 

Topic 08 focuses on Foundation Model Deployment Strategies 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 corporate knowledge assistant must serve intermittent inference requests with minimal always-on application infrastructure. Which approach is the strongest fit?

1. model cascade deployment
2. GPU-optimized container deployment
3. Lambda-based on-demand model invocation
4. Amazon Bedrock provisioned throughput

Correct Answer: C

 

Correct Answer

Answer C is correct because Lambda-based on-demand model invocation is designed to serve intermittent inference requests with minimal always-on application infrastructure. It invokes managed model APIs from serverless code when requests arrive and scales application glue automatically.

Incorrect Answers

Answer A is incorrect because model cascade deployment is primarily used to route easy requests to cheaper models and escalate difficult requests when necessary, which is a different requirement from the one being tested.

Answer B is incorrect because GPU-optimized container deployment is primarily used to serve a large model that requires explicit control over accelerator memory and model runtime, which is a different requirement from the one being tested.

Answer D is incorrect because Amazon Bedrock provisioned throughput is primarily used to reserve predictable model inference capacity for steady or latency-sensitive production demand, which is a different requirement from the one being tested.

 

Question 2

The enterprise data marketplace must reserve predictable model inference capacity for steady or latency-sensitive production demand. Which approach is the strongest fit?

1. endpoint auto scaling
2. Amazon Bedrock provisioned throughput
3. SageMaker AI real-time endpoint
4. optimized model loading strategy

Correct Answer: B

 

Correct Answer

Answer B is correct because Amazon Bedrock provisioned throughput is designed to reserve predictable model inference capacity for steady or latency-sensitive production demand. It allocates dedicated throughput for supported models instead of relying solely on shared on-demand capacity.

Incorrect Answers

Answer A is incorrect because endpoint auto scaling is primarily used to adjust serving capacity as sustained inference demand rises or falls, which is a different requirement from the one being tested.

Answer C is incorrect because SageMaker AI real-time endpoint is primarily used to host a customized or self-managed model artifact behind a managed online inference API, which is a different requirement from the one being tested.

Answer D is incorrect because optimized model loading strategy is primarily used to reduce startup time and memory pressure when deploying large language models, which is a different requirement from the one being tested.

 

Question 3

The case management platform has a design goal to host a customized or self-managed model artifact behind a managed online inference API. What should the team choose?

1. GPU-optimized container deployment
2. smaller pre-trained model for a narrow task
3. SageMaker AI real-time endpoint
4. versioned deployment with rollback

Correct Answer: C

 

Correct Answer

Answer C is correct because SageMaker AI real-time endpoint is designed to host a customized or self-managed model artifact behind a managed online inference API. It provides scalable endpoint infrastructure, deployment configuration, and monitoring for custom models.

Incorrect Answers

Answer A is incorrect because GPU-optimized container deployment is primarily used to serve a large model that requires explicit control over accelerator memory and model runtime, which is a different requirement from the one being tested.

Answer B is incorrect because smaller pre-trained model for a narrow task is primarily used to meet a focused requirement with lower latency and resource cost than a general large model, which is a different requirement from the one being tested.

Answer D is incorrect because versioned deployment with rollback is primarily used to release a new model version safely while preserving a rapid recovery path, which is a different requirement from the one being tested.

 

Question 4

The legal research platform has a design goal to serve a large model that requires explicit control over accelerator memory and model runtime. What should the team choose?

1. GPU-optimized container deployment
2. model cascade deployment
3. optimized model loading strategy
4. hybrid Bedrock and SageMaker deployment

Correct Answer: A

 

Correct Answer

Answer A is correct because GPU-optimized container deployment is designed to serve a large model that requires explicit control over accelerator memory and model runtime. It packages model-serving software in a container sized and configured for gpu utilization and large model memory needs.

Incorrect Answers

Answer B is incorrect because model cascade deployment is primarily used to route easy requests to cheaper models and escalate difficult requests when necessary, which is a different requirement from the one being tested.

Answer C is incorrect because optimized model loading strategy is primarily used to reduce startup time and memory pressure when deploying large language models, which is a different requirement from the one being tested.

Answer D is incorrect because hybrid Bedrock and SageMaker deployment is primarily used to use managed foundation models alongside custom hosted models in one application, which is a different requirement from the one being tested.

 

Question 5

The research assistant application needs an architecture that can reduce startup time and memory pressure when deploying large language models. Which choice best meets that need?

1. optimized model loading strategy
2. smaller pre-trained model for a narrow task
3. endpoint auto scaling
4. Lambda-based on-demand model invocation

Correct Answer: A

 

Correct Answer

Answer A is correct because optimized model loading strategy is designed to reduce startup time and memory pressure when deploying large language models. It uses model-aware loading, caching, quantization, or artifact placement techniques appropriate to the serving stack.

Incorrect Answers

Answer B is incorrect because smaller pre-trained model for a narrow task is primarily used to meet a focused requirement with lower latency and resource cost than a general large model, which is a different requirement from the one being tested.

Answer C is incorrect because endpoint auto scaling is primarily used to adjust serving capacity as sustained inference demand rises or falls, which is a different requirement from the one being tested.

Answer D is incorrect because Lambda-based on-demand model invocation is primarily used to serve intermittent inference requests with minimal always-on application infrastructure, which is a different requirement from the one being tested.

 

Question 6

The AI product engineering team has a design goal to meet a focused requirement with lower latency and resource cost than a general large model. What should the team choose?

1. versioned deployment with rollback
2. smaller pre-trained model for a narrow task
3. Amazon Bedrock provisioned throughput
4. model cascade deployment

Correct Answer: B

 

Correct Answer

Answer B is correct because smaller pre-trained model for a narrow task is designed to meet a focused requirement with lower latency and resource cost than a general large model. It chooses a compact model when evaluation shows it satisfies the quality threshold for the specific task.

Incorrect Answers

Answer A is incorrect because versioned deployment with rollback is primarily used to release a new model version safely while preserving a rapid recovery path, which is a different requirement from the one being tested.

Answer C is incorrect because Amazon Bedrock provisioned throughput is primarily used to reserve predictable model inference capacity for steady or latency-sensitive production demand, which is a different requirement from the one being tested.

Answer D is incorrect because model cascade deployment is primarily used to route easy requests to cheaper models and escalate difficult requests when necessary, which is a different requirement from the one being tested.

 

Question 7

The supply-chain analytics group must route easy requests to cheaper models and escalate difficult requests when necessary. Which approach is the strongest fit?

1. hybrid Bedrock and SageMaker deployment
2. model cascade deployment
3. SageMaker AI real-time endpoint
4. endpoint auto scaling

Correct Answer: B

 

Correct Answer

Answer B is correct because model cascade deployment is designed to route easy requests to cheaper models and escalate difficult requests when necessary. It combines multiple deployed models with decision logic that balances cost, latency, and response quality.

Incorrect Answers

Answer A is incorrect because hybrid Bedrock and SageMaker deployment is primarily used to use managed foundation models alongside custom hosted models in one application, which is a different requirement from the one being tested.

Answer C is incorrect because SageMaker AI real-time endpoint is primarily used to host a customized or self-managed model artifact behind a managed online inference API, which is a different requirement from the one being tested.

Answer D is incorrect because endpoint auto scaling is primarily used to adjust serving capacity as sustained inference demand rises or falls, which is a different requirement from the one being tested.

 

Question 8

The enterprise developer platform is prioritizing a requirement to adjust serving capacity as sustained inference demand rises or falls. Which implementation is most appropriate?

1. endpoint auto scaling
2. Lambda-based on-demand model invocation
3. versioned deployment with rollback
4. GPU-optimized container deployment

Correct Answer: A

 

Correct Answer

Answer A is correct because endpoint auto scaling is designed to adjust serving capacity as sustained inference demand rises or falls. It changes endpoint instance capacity based on observed utilization or request metrics instead of fixing capacity permanently.

Incorrect Answers

Answer B is incorrect because Lambda-based on-demand model invocation is primarily used to serve intermittent inference requests with minimal always-on application infrastructure, which is a different requirement from the one being tested.

Answer C is incorrect because versioned deployment with rollback is primarily used to release a new model version safely while preserving a rapid recovery path, which is a different requirement from the one being tested.

Answer D is incorrect because GPU-optimized container deployment is primarily used to serve a large model that requires explicit control over accelerator memory and model runtime, which is a different requirement from the one being tested.

 

Question 9

The customer identity platform is prioritizing a requirement to release a new model version safely while preserving a rapid recovery path. Which implementation is most appropriate?

1. versioned deployment with rollback
2. Amazon Bedrock provisioned throughput
3. optimized model loading strategy
4. hybrid Bedrock and SageMaker deployment

Correct Answer: A

 

Correct Answer

Answer A is correct because versioned deployment with rollback is designed to release a new model version safely while preserving a rapid recovery path. It keeps previous artifacts and deployment configuration available so failed changes can be reverted predictably.

Incorrect Answers

Answer B is incorrect because Amazon Bedrock provisioned throughput is primarily used to reserve predictable model inference capacity for steady or latency-sensitive production demand, which is a different requirement from the one being tested.

Answer C is incorrect because optimized model loading strategy is primarily used to reduce startup time and memory pressure when deploying large language models, which is a different requirement from the one being tested.

Answer D is incorrect because hybrid Bedrock and SageMaker deployment is primarily used to use managed foundation models alongside custom hosted models in one application, which is a different requirement from the one being tested.

 

Question 10

The business intelligence application is prioritizing a requirement to use managed foundation models alongside custom hosted models in one application. Which implementation is most appropriate?

1. smaller pre-trained model for a narrow task
2. Lambda-based on-demand model invocation
3. SageMaker AI real-time endpoint
4. hybrid Bedrock and SageMaker deployment

Correct Answer: D

 

Correct Answer

Answer D is correct because hybrid Bedrock and SageMaker deployment is designed to use managed foundation models alongside custom hosted models in one application. It routes requests between Bedrock APIs and sagemaker endpoints according to capability, policy, or customization requirements.

Incorrect Answers

Answer A is incorrect because smaller pre-trained model for a narrow task is primarily used to meet a focused requirement with lower latency and resource cost than a general large model, which is a different requirement from the one being tested.

Answer B is incorrect because Lambda-based on-demand model invocation is primarily used to serve intermittent inference requests with minimal always-on application infrastructure, which is a different requirement from the one being tested.

Answer C is incorrect because SageMaker AI real-time endpoint is primarily used to host a customized or self-managed model artifact behind a managed online inference API, which is a different requirement from the one being tested.

 

Question 11

The field analytics service documents this GenAI behavior: invokes managed model APIs from serverless code when requests arrive and scales application glue automatically. Which capability matches it?

1. GPU-optimized container deployment
2. model cascade deployment
3. Amazon Bedrock provisioned throughput
4. Lambda-based on-demand model invocation

Correct Answer: D

 

Correct Answer

Answer D is correct because the description directly matches Lambda-based on-demand model invocation. Lambda-based on-demand model invocation invokes managed model APIs from serverless code when requests arrive and scales application glue automatically.

Incorrect Answers

Answer A is incorrect because GPU-optimized container deployment is primarily used to serve a large model that requires explicit control over accelerator memory and model runtime, which is a different requirement from the one being tested.

Answer B is incorrect because model cascade deployment is primarily used to route easy requests to cheaper models and escalate difficult requests when necessary, which is a different requirement from the one being tested.

Answer C is incorrect because Amazon Bedrock provisioned throughput is primarily used to reserve predictable model inference capacity for steady or latency-sensitive production demand, which is a different requirement from the one being tested.

 

Question 12

The operations intelligence team needs to identify a capability with this behavior: allocates dedicated throughput for supported models instead of relying solely on shared on-demand capacity. What is the best match?

1. Amazon Bedrock provisioned throughput
2. SageMaker AI real-time endpoint
3. optimized model loading strategy
4. endpoint auto scaling

Correct Answer: A

 

Correct Answer

Answer A is correct because the description directly matches Amazon Bedrock provisioned throughput. Amazon Bedrock provisioned throughput allocates dedicated throughput for supported models instead of relying solely on shared on-demand capacity.

Incorrect Answers

Answer B is incorrect because SageMaker AI real-time endpoint is primarily used to host a customized or self-managed model artifact behind a managed online inference API, which is a different requirement from the one being tested.

Answer C is incorrect because optimized model loading strategy is primarily used to reduce startup time and memory pressure when deploying large language models, which is a different requirement from the one being tested.

Answer D is incorrect because endpoint auto scaling is primarily used to adjust serving capacity as sustained inference demand rises or falls, which is a different requirement from the one being tested.

 

Question 13

Within the SaaS product team’s architecture, which capability matches this technical description: provides scalable endpoint infrastructure, deployment configuration, and monitoring for custom models?

1. GPU-optimized container deployment
2. smaller pre-trained model for a narrow task
3. versioned deployment with rollback
4. SageMaker AI real-time endpoint

Correct Answer: D

 

Correct Answer

Answer D is correct because the description directly matches SageMaker AI real-time endpoint. SageMaker AI real-time endpoint provides scalable endpoint infrastructure, deployment configuration, and monitoring for custom models.

Incorrect Answers

Answer A is incorrect because GPU-optimized container deployment is primarily used to serve a large model that requires explicit control over accelerator memory and model runtime, which is a different requirement from the one being tested.

Answer B is incorrect because smaller pre-trained model for a narrow task is primarily used to meet a focused requirement with lower latency and resource cost than a general large model, which is a different requirement from the one being tested.

Answer C is incorrect because versioned deployment with rollback is primarily used to release a new model version safely while preserving a rapid recovery path, which is a different requirement from the one being tested.

 

Question 14

Within the customer-service platform’s architecture, which capability matches this technical description: packages model-serving software in a container sized and configured for GPU utilization and large model memory needs?

1. hybrid Bedrock and SageMaker deployment
2. GPU-optimized container deployment
3. model cascade deployment
4. optimized model loading strategy

Correct Answer: B

 

Correct Answer

Answer B is correct because the description directly matches GPU-optimized container deployment. GPU-optimized container deployment packages model-serving software in a container sized and configured for GPU utilization and large model memory needs.

Incorrect Answers

Answer A is incorrect because hybrid Bedrock and SageMaker deployment is primarily used to use managed foundation models alongside custom hosted models in one application, which is a different requirement from the one being tested.

Answer C is incorrect because model cascade deployment is primarily used to route easy requests to cheaper models and escalate difficult requests when necessary, which is a different requirement from the one being tested.

Answer D is incorrect because optimized model loading strategy is primarily used to reduce startup time and memory pressure when deploying large language models, which is a different requirement from the one being tested.

 

Question 15

The cloud center of excellence documents this GenAI behavior: uses model-aware loading, caching, quantization, or artifact placement techniques appropriate to the serving stack. Which capability matches it?

1. smaller pre-trained model for a narrow task
2. endpoint auto scaling
3. Lambda-based on-demand model invocation
4. optimized model loading strategy

Correct Answer: D

 

Correct Answer

Answer D is correct because the description directly matches optimized model loading strategy. optimized model loading strategy uses model-aware loading, caching, quantization, or artifact placement techniques appropriate to the serving stack.

Incorrect Answers

Answer A is incorrect because smaller pre-trained model for a narrow task is primarily used to meet a focused requirement with lower latency and resource cost than a general large model, which is a different requirement from the one being tested.

Answer B is incorrect because endpoint auto scaling is primarily used to adjust serving capacity as sustained inference demand rises or falls, which is a different requirement from the one being tested.

Answer C is incorrect because Lambda-based on-demand model invocation is primarily used to serve intermittent inference requests with minimal always-on application infrastructure, which is a different requirement from the one being tested.

 

Question 16

The contact-center transformation team documents this GenAI behavior: chooses a compact model when evaluation shows it satisfies the quality threshold for the specific task. Which capability matches it?

1. smaller pre-trained model for a narrow task
2. model cascade deployment
3. Amazon Bedrock provisioned throughput
4. versioned deployment with rollback

Correct Answer: A

 

Correct Answer

Answer A is correct because the description directly matches smaller pre-trained model for a narrow task. smaller pre-trained model for a narrow task chooses a compact model when evaluation shows it satisfies the quality threshold for the specific task.

Incorrect Answers

Answer B is incorrect because model cascade deployment is primarily used to route easy requests to cheaper models and escalate difficult requests when necessary, which is a different requirement from the one being tested.

Answer C is incorrect because Amazon Bedrock provisioned throughput is primarily used to reserve predictable model inference capacity for steady or latency-sensitive production demand, which is a different requirement from the one being tested.

Answer D is incorrect because versioned deployment with rollback is primarily used to release a new model version safely while preserving a rapid recovery path, which is a different requirement from the one being tested.

 

Question 17

The customer analytics platform describes a component that combines multiple deployed models with decision logic that balances cost, latency, and response quality. Which capability is being described?

1. hybrid Bedrock and SageMaker deployment
2. SageMaker AI real-time endpoint
3. model cascade deployment
4. endpoint auto scaling

Correct Answer: C

 

Correct Answer

Answer C is correct because the description directly matches model cascade deployment. model cascade deployment combines multiple deployed models with decision logic that balances cost, latency, and response quality.

Incorrect Answers

Answer A is incorrect because hybrid Bedrock and SageMaker deployment is primarily used to use managed foundation models alongside custom hosted models in one application, which is a different requirement from the one being tested.

Answer B is incorrect because SageMaker AI real-time endpoint is primarily used to host a customized or self-managed model artifact behind a managed online inference API, which is a different requirement from the one being tested.

Answer D is incorrect because endpoint auto scaling is primarily used to adjust serving capacity as sustained inference demand rises or falls, which is a different requirement from the one being tested.

 

Question 18

The enterprise finance automation team documents this GenAI behavior: changes endpoint instance capacity based on observed utilization or request metrics instead of fixing capacity permanently. Which capability matches it?

1. Lambda-based on-demand model invocation
2. endpoint auto scaling
3. GPU-optimized container deployment
4. versioned deployment with rollback

Correct Answer: B

 

Correct Answer

Answer B is correct because the description directly matches endpoint auto scaling. endpoint auto scaling changes endpoint instance capacity based on observed utilization or request metrics instead of fixing capacity permanently.

Incorrect Answers

Answer A is incorrect because Lambda-based on-demand model invocation is primarily used to serve intermittent inference requests with minimal always-on application infrastructure, which is a different requirement from the one being tested.

Answer C is incorrect because GPU-optimized container deployment is primarily used to serve a large model that requires explicit control over accelerator memory and model runtime, which is a different requirement from the one being tested.

Answer D is incorrect because versioned deployment with rollback is primarily used to release a new model version safely while preserving a rapid recovery path, which is a different requirement from the one being tested.

 

Question 19

The risk and controls team describes a component that keeps previous artifacts and deployment configuration available so failed changes can be reverted predictably. Which capability is being described?

1. hybrid Bedrock and SageMaker deployment
2. versioned deployment with rollback
3. optimized model loading strategy
4. Amazon Bedrock provisioned throughput

Correct Answer: B

 

Correct Answer

Answer B is correct because the description directly matches versioned deployment with rollback. versioned deployment with rollback keeps previous artifacts and deployment configuration available so failed changes can be reverted predictably.

Incorrect Answers

Answer A is incorrect because hybrid Bedrock and SageMaker deployment is primarily used to use managed foundation models alongside custom hosted models in one application, which is a different requirement from the one being tested.

Answer C is incorrect because optimized model loading strategy is primarily used to reduce startup time and memory pressure when deploying large language models, which is a different requirement from the one being tested.

Answer D is incorrect because Amazon Bedrock provisioned throughput is primarily used to reserve predictable model inference capacity for steady or latency-sensitive production demand, which is a different requirement from the one being tested.

 

Question 20

Within the enterprise search architecture group’s architecture, which capability matches this technical description: routes requests between Bedrock APIs and SageMaker endpoints according to capability, policy, or customization requirements?

1. hybrid Bedrock and SageMaker deployment
2. Lambda-based on-demand model invocation
3. smaller pre-trained model for a narrow task
4. SageMaker AI real-time endpoint

Correct Answer: A

 

Correct Answer

Answer A is correct because the description directly matches hybrid Bedrock and SageMaker deployment. hybrid Bedrock and SageMaker deployment routes requests between Bedrock APIs and SageMaker endpoints according to capability, policy, or customization requirements.

Incorrect Answers

Answer B is incorrect because Lambda-based on-demand model invocation is primarily used to serve intermittent inference requests with minimal always-on application infrastructure, which is a different requirement from the one being tested.

Answer C is incorrect because smaller pre-trained model for a narrow task is primarily used to meet a focused requirement with lower latency and resource cost than a general large model, which is a different requirement from the one being tested.

Answer D is incorrect because SageMaker AI real-time endpoint is primarily used to host a customized or self-managed model artifact behind a managed online inference API, which is a different requirement from the one being tested.

 

Question 21

Why would the service reliability organization introduce Lambda-based on-demand model invocation into the GenAI architecture?

1. Serve a large model that requires explicit control over accelerator memory and model runtime
2. Reserve predictable model inference capacity for steady or latency-sensitive production demand
3. Route easy requests to cheaper models and escalate difficult requests when necessary
4. Serve intermittent inference requests with minimal always-on application infrastructure

Correct Answer: D

 

Correct Answer

Answer D is correct because Lambda-based on-demand model invocation is specifically used to serve intermittent inference requests with minimal always-on application infrastructure. It invokes managed model APIs from serverless code when requests arrive and scales application glue automatically.

Incorrect Answers

Answer A is incorrect because that requirement aligns with GPU-optimized container deployment, not Lambda-based on-demand model invocation.

Answer B is incorrect because that requirement aligns with Amazon Bedrock provisioned throughput, not Lambda-based on-demand model invocation.

Answer C is incorrect because that requirement aligns with model cascade deployment, not Lambda-based on-demand model invocation.

 

Question 22

The financial-services engineering group plans to adopt Amazon Bedrock provisioned throughput. Which outcome should drive that decision?

1. Reduce startup time and memory pressure when deploying large language models
2. Host a customized or self-managed model artifact behind a managed online inference API
3. Adjust serving capacity as sustained inference demand rises or falls
4. Reserve predictable model inference capacity for steady or latency-sensitive production demand

Correct Answer: D

 

Correct Answer

Answer D is correct because Amazon Bedrock provisioned throughput is specifically used to reserve predictable model inference capacity for steady or latency-sensitive production demand. It allocates dedicated throughput for supported models instead of relying solely on shared on-demand capacity.

Incorrect Answers

Answer A is incorrect because that requirement aligns with optimized model loading strategy, not Amazon Bedrock provisioned throughput.

Answer B is incorrect because that requirement aligns with SageMaker AI real-time endpoint, not Amazon Bedrock provisioned throughput.

Answer C is incorrect because that requirement aligns with endpoint auto scaling, not Amazon Bedrock provisioned throughput.

 

Question 23

The software delivery organization has proposed SageMaker AI real-time endpoint for its design. Which requirement best justifies it?

1. Meet a focused requirement with lower latency and resource cost than a general large model
2. Release a new model version safely while preserving a rapid recovery path
3. Host a customized or self-managed model artifact behind a managed online inference API
4. Serve a large model that requires explicit control over accelerator memory and model runtime

Correct Answer: C

 

Correct Answer

Answer C is correct because SageMaker AI real-time endpoint is specifically used to host a customized or self-managed model artifact behind a managed online inference API. It provides scalable endpoint infrastructure, deployment configuration, and monitoring for custom models.

Incorrect Answers

Answer A is incorrect because that requirement aligns with smaller pre-trained model for a narrow task, not SageMaker AI real-time endpoint.

Answer B is incorrect because that requirement aligns with versioned deployment with rollback, not SageMaker AI real-time endpoint.

Answer D is incorrect because that requirement aligns with GPU-optimized container deployment, not SageMaker AI real-time endpoint.

 

Question 24

Within the claims processing service’s design, the team highlights GPU-optimized container deployment. Which need does that component address?

1. Route easy requests to cheaper models and escalate difficult requests when necessary
2. Use managed foundation models alongside custom hosted models in one application
3. Serve a large model that requires explicit control over accelerator memory and model runtime
4. Reduce startup time and memory pressure when deploying large language models

Correct Answer: C

 

Correct Answer

Answer C is correct because GPU-optimized container deployment is specifically used to serve a large model that requires explicit control over accelerator memory and model runtime. It packages model-serving software in a container sized and configured for gpu utilization and large model memory needs.

Incorrect Answers

Answer A is incorrect because that requirement aligns with model cascade deployment, not GPU-optimized container deployment.

Answer B is incorrect because that requirement aligns with hybrid Bedrock and SageMaker deployment, not GPU-optimized container deployment.

Answer D is incorrect because that requirement aligns with optimized model loading strategy, not GPU-optimized container deployment.

 

Question 25

The clinical knowledge application plans to adopt optimized model loading strategy. Which outcome should drive that decision?

1. Meet a focused requirement with lower latency and resource cost than a general large model
2. Serve intermittent inference requests with minimal always-on application infrastructure
3. Reduce startup time and memory pressure when deploying large language models
4. Adjust serving capacity as sustained inference demand rises or falls

Correct Answer: C

 

Correct Answer

Answer C is correct because optimized model loading strategy is specifically used to reduce startup time and memory pressure when deploying large language models. It uses model-aware loading, caching, quantization, or artifact placement techniques appropriate to the serving stack.

Incorrect Answers

Answer A is incorrect because that requirement aligns with smaller pre-trained model for a narrow task, not optimized model loading strategy.

Answer B is incorrect because that requirement aligns with Lambda-based on-demand model invocation, not optimized model loading strategy.

Answer D is incorrect because that requirement aligns with endpoint auto scaling, not optimized model loading strategy.