AIP-C01 Practice Test 5 — 25 Questions

Question 1

A financial services company is deploying a generative AI application that uses Amazon Bedrock to answer employee questions about internal policies. The security team requires that the application automatically redact personally identifiable information (PII) such as account numbers and social security numbers from both user prompts and model responses, without modifying the underlying foundation model. Which AWS service should be integrated into the application's request and response pipeline to meet this requirement?

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

Amazon Comprehend. Amazon Comprehend offers purpose-built PII detection and redaction APIs that can be invoked synchronously within an application's request and response flow. Integrating it before Bedrock invocation and after completion generation ensures sensitive data is masked without retraining or modifying the foundation model. The other services address storage classification, audit logging, or threat detection, none of which provide inline content redaction for generative AI payloads.

Answer

Amazon Macie

Answer

AWS CloudTrail

Answer

Amazon GuardDuty

Question 2

A company is deploying a generative AI application on AWS that uses Amazon Bedrock to generate product descriptions. The application must ensure that the generated text does not contain any personally identifiable information (PII) or offensive content. The security team requires that all generated content be automatically scanned and filtered before being stored in Amazon S3. Which AWS service or feature should be integrated into the application pipeline to meet these requirements?

Accepted Answer

Amazon Bedrock Guardrails. Amazon Bedrock Guardrails is purpose-built to apply safeguards to generative AI outputs, including PII redaction and content filtering. It integrates seamlessly with Bedrock, allowing automatic scanning and blocking of undesired content before it is stored or returned. Other services like Comprehend, Lambda, or Macie are not designed for this real-time, integrated filtering within the Bedrock workflow.

Answer

Amazon Macie

Answer

Amazon Comprehend

Answer

AWS Lambda with custom regex patterns

Question 3

A developer is using Amazon Bedrock to build a chatbot that can answer questions about a company's internal documents. The developer needs to implement a retrieval-augmented generation (RAG) workflow. Which AWS service should be used to store document embeddings and perform similarity search?

Accepted Answer

Amazon OpenSearch Serverless with vector engine. Amazon OpenSearch Serverless includes a vector engine specifically designed for storing and querying embeddings, making it ideal for RAG applications. It integrates with Amazon Bedrock Knowledge Bases to provide managed retrieval. Other options like RDS, DynamoDB, or S3 lack native vector search capabilities or are not optimized for this purpose.

Answer

Amazon DynamoDB

Answer

Amazon RDS for PostgreSQL

Answer

Amazon S3 with Glacier Deep Archive

Question 4

A generative AI team runs nightly batch inference on Amazon Bedrock for 40 million prompts. Each job currently invokes the model once per prompt with a synchronous Converse call and the job takes nine hours, exceeding the six-hour window. The prompts are independent and the team needs the same model quality. Which change most directly reduces wall-clock time while preserving per-prompt output quality?

Accepted Answer

Submit the prompts through the Bedrock batch inference job API and poll for job completion.. Batch inference is purpose-built for high-volume, latency-tolerant generative AI workloads on Bedrock. It ingests a JSONL manifest from Amazon S3, executes the same foundation model asynchronously, and returns outputs to S3, so quality is unchanged while throughput scales far beyond what a synchronous per-prompt loop can achieve. It also avoids client-side retry storms and quota contention that plague naive looping.

Answer

Switch to a smaller, cheaper model for the entire batch and accept lower quality.

Answer

Increase the InvokeModel requests-per-minute quota and keep the same synchronous loop.

Answer

Move the loop to an AWS Lambda function with a 15-minute timeout and higher memory.

Question 5

A healthcare startup is preparing a generative AI chatbot that answers patient questions using Amazon Bedrock. The security team wants to ensure that protected health information is not inadvertently echoed back in model responses and that the model cannot be prompted to reveal its system instructions. Which AWS feature should they enable first to address both concerns?

Accepted Answer

Amazon Bedrock Guardrails with sensitive information filters and prompt attack filters.. Amazon Bedrock Guardrails is the purpose-built control for inspecting model inputs and outputs. Sensitive information filters detect and mask or block PHI and PII, while prompt attack filters identify jailbreak and prompt-injection patterns that try to extract system prompts. Enabling Guardrails directly addresses both the PHI disclosure and system instruction leakage concerns.

Answer

Amazon VPC endpoints for Bedrock to keep traffic off the public internet.

Answer

AWS Key Management Service customer managed keys for the Bedrock model customization jobs.

Answer

AWS Identity and Access Management policies that restrict which Bedrock models the application can invoke.

Question 6

A developer is testing a retrieval-augmented generation (RAG) application built on Amazon Bedrock Knowledge Bases. For a specific question, the model returns a confident but incorrect answer, and the retrieved chunks shown in the test console are only loosely related to the question. The developer wants to diagnose whether the retrieval step or the generation step is at fault. Which approach is most effective?

Accepted Answer

Use the Retrieve API to fetch chunks for the question and inspect their relevance scores and content independently of generation.. To separate retrieval failures from generation failures, you must inspect the retrieved chunks independently. The Retrieve API returns the chunks and relevance scores without invoking the generation model, so you can verify whether the correct context was available. If the right context is present but the answer is wrong, generation is at fault; otherwise, retrieval tuning is needed.

Answer

Add a guardrail that blocks unsupported answers and observe whether the response is filtered, indicating a generation problem.

Answer

Switch the knowledge base to a different embedding model and re-ingest all documents, then compare the final answers.

Answer

Increase the model's temperature to 1.0 and re-run the query to see if the answer changes, then compare outputs.

Question 7

A financial services firm is building a RAG application using Amazon Bedrock Knowledge Bases. They need to ensure that the vector embeddings for their proprietary documents are stored in a way that meets strict data residency requirements. The documents are stored in an Amazon S3 bucket in the eu-west-1 region. They want to use Amazon OpenSearch Serverless as the vector store. Which configuration should they choose?

Accepted Answer

Create the OpenSearch Serverless collection in eu-west-1 and configure the Bedrock Knowledge Base to use it.. Data residency requires that all data, including vector embeddings, remain within the specified region. By creating an OpenSearch Serverless collection in eu-west-1 and configuring the Bedrock Knowledge Base to use it, the embeddings are stored and processed within the EU. Other options either place data in a different region or use an alternative vector store that does not match the stated preference.

Answer

Create an OpenSearch Serverless collection in us-east-1 and use VPC peering to connect to the S3 bucket in eu-west-1.

Answer

Use the default OpenSearch Serverless collection provided by Bedrock Knowledge Bases, which automatically replicates data across regions.

Answer

Use Amazon Aurora PostgreSQL with pgvector as the vector store, deployed in eu-west-1, and configure the Knowledge Base to use it.

Question 8

A developer is building a document summarization feature on Amazon Bedrock. The same 30-page policy document is summarized by hundreds of users each day, and the document changes only when legal publishes a new revision. The summaries are currently regenerated per request, and the team wants to cut token spend without changing the summary text. Which approach best meets the goal?

Accepted Answer

Store the generated summary in Amazon S3 keyed by document revision and serve it on repeat requests.. When the source document is stable across a revision, the summary is effectively a deterministic artifact. Materializing it once in S3 and keying it by document revision turns hundreds of daily generations into a single generation plus cheap object reads. This removes the dominant token cost while guaranteeing users see identical text, which is exactly what the team asked for.

Answer

Increase the max tokens parameter so the summary completes in one pass.

Answer

Lower the model temperature to zero and keep regenerating the summary per request.

Answer

Enable prompt caching on the Bedrock model for the document prefix.

Question 9

A developer is building a generative AI application using Amazon Bedrock and needs to ensure that all API calls to the Bedrock runtime are logged for auditing purposes. The logs must include the request and response payloads, and must be retained for 90 days. The developer wants to minimize operational overhead. Which solution meets these requirements?

Accepted Answer

Enable model invocation logging in Amazon Bedrock, configure it to log to Amazon CloudWatch Logs, and set the retention period to 90 days.. Model invocation logging is a native Bedrock feature that can capture full request and response data and deliver it to CloudWatch Logs or S3. Configuring it to CloudWatch Logs with a 90-day retention period is the simplest way to meet the auditing requirement with minimal overhead. Other options either do not capture payloads or require custom code.

Answer

Create an AWS Lambda function that intercepts all Bedrock API calls, logs the payloads to Amazon S3, and applies a 90-day lifecycle policy.

Answer

Use AWS CloudTrail to log all Bedrock API calls, and configure an S3 bucket with a lifecycle policy to delete objects after 90 days.

Answer

Enable VPC Flow Logs for the VPC endpoint used by Bedrock, and store logs in CloudWatch with a 90-day retention period.

Question 10

A developer is testing an Amazon Bedrock application and wants to verify that the model's responses do not contain personally identifiable information (PII) before the application displays them to users. The developer needs a managed way to detect and block PII in both prompts and responses. Which AWS feature should be used?

Accepted Answer

Amazon Bedrock Guardrails with the sensitive information filter configured for PII.. Amazon Bedrock Guardrails include a sensitive information filter that can detect and block or mask PII in both input prompts and model responses. Configuring this filter enforces PII policy at inference time without custom code, satisfying the need for a managed solution that covers both directions of the interaction.

Answer

Amazon Comprehend PII detection called manually after each inference to filter outputs.

Answer

AWS WAF with a custom rule to inspect request and response bodies for PII patterns.

Answer

Amazon Macie to scan the model's responses for PII before display.

Question 11

Which TWO types of models are generally available in Amazon Bedrock?

Accepted Answer

Embedding models.. Amazon Bedrock provides foundation models (FMs) accessed via a unified API, and these FMs fall into categories such as embedding models (Option A) and text generation models (Option E). Embedding models in Bedrock, such as Amazon Titan Embeddings or Cohere Embed, convert text into vector representations used for semantic search, RAG, and clustering, which is why Option A is correct. Text generation models, such as Anthropic Claude, Amazon Titan Text, Meta Llama, and Mistral, produce natural-language output for chat, summarization, and content creation, which is why Option E is correct. Options B, C, and D are incorrect because network load balancer configurations, hardware drivers, and OS kernel images are infrastructure and systems-level artifacts, not model types offered by Amazon Bedrock.

Answer

Network load balancer configurations.

Answer

Hardware drivers.

Answer

OS kernel images.

Question 12

A developer is building a RAG application using Amazon Bedrock Knowledge Bases. The knowledge base is configured with an Amazon OpenSearch Serverless vector store. They notice that queries are returning irrelevant results. They suspect that the chunking strategy is not optimal for their documents, which are technical manuals with long paragraphs and code snippets. Which action should they take to improve retrieval accuracy?

Accepted Answer

Switch to a hierarchical chunking strategy that preserves document structure.. Hierarchical chunking preserves the structure of technical documents by creating parent and child chunks, allowing the retriever to match queries to precise child chunks while still providing the broader parent context. This is particularly effective for manuals with code snippets and long paragraphs, as it maintains semantic coherence and improves retrieval accuracy.

Answer

Reduce the number of retrieved chunks by lowering the top-k parameter.

Answer

Increase the chunk size to the maximum allowed by Bedrock Knowledge Bases.

Answer

Enable hybrid search by combining keyword and vector search.

Question 13

A company runs a customer-facing chatbot on Amazon Bedrock with provisioned throughput purchased as a model unit for a specific foundation model in us-east-1. Traffic is steady during business hours but drops to near zero overnight, and the team is paying for the commitment around the clock. Leadership wants to stop paying for idle overnight capacity without risking daytime throttling. Which action best fits?

Accepted Answer

Model the daytime steady load as on-demand and reserve provisioned throughput only if sustained utilization justifies the commitment.. Provisioned throughput is a monthly commitment that guarantees capacity but bills around the clock, so it only pays off when sustained utilization is high enough to beat on-demand rates. With traffic collapsing overnight, the commitment charges for idle hours. Right-sizing by serving steady load on-demand and reserving only when a utilization analysis shows a durable break-even point removes the idle spend while keeping daytime capacity adequate.

Answer

Move the workload to a smaller foundation model so the existing model unit covers more traffic.

Answer

Keep provisioned throughput sized for the overnight trough and rely on on-demand for daytime peaks.

Answer

Delete the provisioned throughput purchase each evening and recreate it each morning.

Question 14

A developer is troubleshooting a generative AI application that calls Amazon Bedrock. Intermittently, invocations fail with a ThrottlingException. CloudWatch metrics show that the account's on-demand quota for the model is being exceeded during peak hours. The developer wants to reduce throttling without changing the model or application logic. Which solution is most appropriate?

Accepted Answer

Implement exponential backoff with jitter in the AWS SDK retry configuration and request a quota increase for the model.. Intermittent ThrottlingException from Bedrock indicates the account is exceeding its on-demand quota. Exponential backoff with jitter allows retries to succeed when throttling is transient, and requesting a quota increase raises the sustained capacity limit. This combination reduces throttling without modifying the model or application logic, addressing both immediate and long-term needs.

Answer

Reduce the maxTokenCount parameter to 1 to lower token consumption per request.

Answer

Switch the application to use a different AWS Region where the model is available and retry all requests there.

Answer

Enable provisioned throughput for the model and remove all retry logic from the application.

Question 15

You notice that your SageMaker model endpoint experiences cold starts. Which feature can mitigate this?

Accepted Answer

Enabling Provisioned Concurrency. Provisioned Concurrency maintains a set number of initialized instances, eliminating cold starts for serverless inference.

Answer

Increasing the instance count

Answer

Changing the container image

Answer

Using a smaller model

Question 16

When invoking models in Bedrock, what does the 'streaming' feature provide?

Accepted Answer

Reduced latency until the first token appears. Streaming in Amazon Bedrock returns the model's output incrementally as tokens are generated, rather than waiting for the full response. The primary benefit is reduced latency until the first token appears, improving perceived responsiveness for interactive applications. It does not change the model's accuracy, context window, or cost.

Answer

Support for larger context windows

Answer

Higher model accuracy

Answer

Lower cost per invocation

Question 17

Which TWO factors are critical when choosing a model for a RAG architecture that requires high grounding accuracy?

Accepted Answer

The model's support for citation and source attribution.. Option C is correct because high grounding accuracy in RAG depends on the model's ability to cite and attribute retrieved sources, letting users verify that answers are anchored in the provided documents rather than hallucinated. Option E is correct because a larger context window allows more retrieved chunks to be included in the prompt, giving the model sufficient evidence to ground its response accurately. Option A is not critical, since the training date affects knowledge freshness but RAG grounding comes from retrieved context, not parametric memory. Option B is irrelevant, as image generation has no bearing on textual grounding accuracy. Option D is not decisive, because parameter count alone does not determine how faithfully a model uses retrieved context.

Answer

The model's training date.

Answer

The model's ability to generate images.

Answer

The number of parameters in the model.

Question 18

A company is building a RAG application using Amazon Bedrock and needs to ensure that the data ingested into the knowledge base is filtered based on user identity. Which approach should the developer use?

Accepted Answer

Use Amazon Bedrock Knowledge Base metadata filtering during the retrieval phase to apply access controls.. Amazon Bedrock Knowledge Bases support metadata filtering during retrieval, allowing the application to pass user identity attributes as filters so only documents the user is authorized to see are retrieved. This enforces access control at the retrieval phase, before the model generates a response. It is the intended mechanism for identity-based filtering in RAG.

Answer

Use Lambda authorizers to filter the retrieved results after they are returned by the model.

Answer

Create separate vector indexes for every user role within the Knowledge Base configuration.

Answer

Implement fine-grained access control using IAM policies on the S3 bucket containing the source data.

Question 19

You need to ensure that your SageMaker inference endpoint does not exceed a specific monthly budget. Which AWS feature allows you to set up notifications when spending thresholds are crossed?

Accepted Answer

AWS Budgets. AWS Budgets allows you to set custom cost and usage budgets and receive email or SNS alerts when thresholds are approached or exceeded.

Answer

SageMaker Cost Explorer

Answer

CloudWatch Alarms on billing metrics

Answer

AWS Cost and Usage Reports

Question 20

Which metric should a developer monitor to ensure the RAG system remains performant as usage increases?

Accepted Answer

Inference Latency.. Latency and Token Consumption (or request count) are key indicators of RAG performance and cost efficiency.

Answer

Vector Database Storage Capacity.

Answer

IAM User Count.

Answer

Model Training Loss.