AIF-C01 Practice Exam 52 — 20 Questions

Question 1

A company is using a foundation model on Amazon Bedrock to generate customer support responses. They notice that the model sometimes produces harmful or offensive content. Which approach is MOST effective to mitigate this issue?

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

Accepted Answer

Configure Amazon Bedrock Guardrails with content filters. Amazon Bedrock Guardrails provides configurable content filters that can block harmful, offensive, or inappropriate content in both user inputs and model outputs. This is the most effective approach because it operates at the inference layer, applying safety policies consistently across all requests without requiring model retraining or manual review. Prompt engineering alone is unreliable, and fine-tuning may not generalize to all harmful content patterns.

Answer

Use prompt engineering to instruct the model to avoid harmful content

Answer

Enable model invocation logging to review and block responses

Answer

Fine-tune the model on a curated dataset of safe responses

Question 2

A company is building a generative AI application to answer questions from a large set of technical manuals. Which TWO services or features can be used together in a RAG architecture on AWS? (Choose TWO.)

Accepted Answer

Amazon OpenSearch Serverless. Amazon OpenSearch Serverless (B) is correct because it provides a vector search collection that can store and retrieve document embeddings, serving as the retrieval layer in a RAG architecture for querying the technical manuals. Amazon Bedrock Knowledge Bases (D) is correct because it is a fully managed RAG feature that ingests source documents, generates embeddings, stores them in a vector store (such as OpenSearch Serverless), and orchestrates retrieval-augmented generation with foundation models. Together, these two services directly implement the retrieval and generation components of a RAG pipeline on AWS. Amazon SageMaker (A) is a general ML platform for building, training, and deploying models, not a purpose-built RAG retrieval or knowledge base service. AWS Lambda (C) is a serverless compute service that could glue components together but is not itself a RAG retrieval or knowledge base feature. Amazon Bedrock Guardrails (E) applies safety and content filtering policies to model inputs and outputs, which is unrelated to the retrieval and knowledge grounding required by RAG.

Answer

Amazon SageMaker

Answer

AWS Lambda

Answer

Amazon Bedrock Guardrails

Question 3

An organization wants to detect anomalies in real-time streaming data from IoT devices. The data includes sensor readings, and the team plans to use a machine learning model. Which AWS service should be used to build and deploy the model with minimal operational overhead?

Accepted Answer

Amazon SageMaker. Amazon SageMaker is the correct choice because it provides a fully managed environment for building, training, and deploying machine learning models at scale. For real-time anomaly detection on streaming IoT data, SageMaker can host a trained model as a real-time endpoint that processes incoming sensor readings via Amazon Kinesis Data Streams or AWS Lambda, minimizing operational overhead by handling infrastructure, scaling, and monitoring automatically.

Answer

AWS Glue

Answer

Amazon QuickSight

Answer

Amazon Kinesis Data Analytics

Question 4

A machine learning team is working on a multi-label classification problem. They have a highly imbalanced dataset where some labels appear very infrequently. Which evaluation metric is MOST appropriate for assessing model performance across all labels?

Accepted Answer

Macro-averaged F1 score. Macro-averaged F1 score is the most appropriate metric for multi-label classification with highly imbalanced data because it computes the F1 score independently for each label and then averages them, giving equal weight to all labels regardless of their frequency. This ensures that the performance on rare labels is not overshadowed by the performance on frequent labels, which is critical when infrequent labels are equally important.

Answer

Precision

Answer

Micro-averaged F1 score

Answer

Accuracy

Question 5

A financial services firm is deploying a generative AI chatbot using Amazon Bedrock. They must ensure that the chatbot does not generate investment advice and that it automatically redacts any personally identifiable information (PII) from user inputs before processing. Which TWO Bedrock features should they use?

Accepted Answer

Bedrock Guardrails with topic denial. Option A is correct because Bedrock Guardrails support topic denial policies, which let you define denied topics (such as investment advice) so the model refuses to generate content on those subjects, directly satisfying the requirement to prevent investment advice. Option C is correct because Bedrock Guardrails provide sensitive information filters with PII detection and redaction, which identify and mask PII in user inputs before processing, meeting the redaction requirement. Option B (Knowledge Bases) is for retrieval-augmented generation over your own data sources and does not enforce content restrictions or redact PII. Option D (Bedrock Studio) is a collaborative development environment for building generative AI applications, not a content-safety or redaction control. Option E (Bedrock Agents) orchestrates multi-step tasks and API calls, but it does not itself deny topics or redact PII.

Answer

Bedrock Knowledge Bases

Answer

Bedrock Studio

Answer

Bedrock Agents

Question 6

A company uses Amazon Bedrock with a third-party foundation model. They are concerned about the third-party provider accessing their data. What should they review to understand data handling practices?

Accepted Answer

The third-party model provider's data privacy and handling documentation within AWS Bedrock's service description. When using a third-party foundation model through Amazon Bedrock, the model provider's own data privacy and handling documentation — surfaced within the Bedrock service description and model details — is the authoritative source for how that provider treats your prompts, completions, and any fine-tuning data. AWS's shared responsibility model means AWS secures the Bedrock infrastructure, but the third-party model provider defines its own data usage terms. Reviewing that documentation tells you whether inputs are used for training, how long they are retained, and what regional or contractual safeguards apply.

Answer

AWS Artifact reports for SOC and PCI compliance

Answer

AWS CloudTrail logs for model invocation

Answer

Amazon SageMaker Model Registry metadata

Question 7

A company operates a customer support chatbot that uses Amazon Bedrock with a knowledge base sourced from an S3 bucket containing frequently updated product documentation. The knowledge base uses OpenSearch Serverless as the vector store and is configured to sync daily. The chatbot uses the RetrieveAndGenerate API with a custom Lambda function that applies a system prompt instructing the model to base answers solely on the retrieved context. After a major update to the product documentation, the IT team verifies that the data source sync completed successfully and the new chunks are present in the OpenSearch index. However, the chatbot continues to respond with outdated information. Further investigation reveals that the Lambda function includes a response caching mechanism using Amazon ElastiCache for Redis with a Time-To-Live (TTL) of 24 hours. The cache key is based on the user query. The team notes that no cache invalidation is performed after documentation updates. What is the most likely cause of the outdated responses?

Accepted Answer

The ElastiCache cache is returning stale cached responses that contain the old information.. The Lambda function caches responses in ElastiCache for Redis with a 24-hour TTL keyed on the user query, and no invalidation occurs after documentation updates. Even though the knowledge base sync succeeded and new chunks are in OpenSearch, the chatbot returns the cached stale answer for any query previously cached. This is the classic cache-staleness failure mode.

Answer

The 'maximum results' parameter in the RetrieveAndGenerate API is set to a value too low to retrieve the new chunks.

Answer

The embedding model used by the knowledge base has not been retrained on the new documentation.

Answer

The IAM role for the Lambda function lacks permissions to access the new S3 objects.

Question 8

A company wants to automatically discover sensitive data such as credit card numbers in their Amazon S3 training datasets before using them for model training. Which AWS service should they use?

Accepted Answer

Amazon Macie. Amazon Macie is a fully managed data security service that uses machine learning and pattern matching to automatically discover, classify, and protect sensitive data stored in Amazon S3. It is purpose-built to detect PII such as credit card numbers, social security numbers, and API keys in S3 buckets, making it the correct choice for scanning training datasets before use.

Answer

AWS Config

Answer

Amazon GuardDuty

Answer

AWS Audit Manager

Question 9

A company is deploying a generative AI application using Amazon Bedrock. The application must comply with data residency requirements, so all data processing must occur within a specific AWS Region. Additionally, the company wants to reduce latency for end users. Which TWO actions should the company take? (Choose TWO.)

Accepted Answer

Use a smaller, faster foundation model. Option C is correct because using a smaller, faster foundation model reduces inference latency, directly addressing the requirement to lower latency for end users. Option D is correct because selecting a foundation model that is available in the target AWS Region ensures all data processing stays within that Region, satisfying the data residency requirement. Options A, B, and E do not belong: a larger model increases latency rather than reducing it, a global Aurora database would replicate data across Regions and violate data residency, and cross-Region inference would route requests outside the target Region, also breaking data residency.

Answer

Use a larger model with more parameters for better accuracy

Answer

Store all embeddings in a global Aurora database

Answer

Enable cross-Region inference to distribute load

Question 10

A developer is building a RAG-based Q&A bot with Amazon Bedrock Knowledge Bases. They need a managed vector store for document embeddings. Which service should they use?

Accepted Answer

Amazon OpenSearch Serverless. Amazon Bedrock Knowledge Bases requires a vector store to store and query document embeddings for Retrieval-Augmented Generation (RAG). Amazon OpenSearch Serverless provides a managed, scalable vector engine that supports k-NN (k-nearest neighbor) search, making it the correct choice for this use case. It integrates natively with Bedrock Knowledge Bases to handle embedding storage and similarity search without manual infrastructure management.

Answer

Amazon DynamoDB

Answer

Amazon RDS

Answer

Amazon S3

Question 11

A machine learning engineer needs to choose a service to extract text from scanned PDF forms, including handwritten fields. Which AWS service is MOST appropriate?

Accepted Answer

Amazon Textract. Amazon Textract is specifically designed to extract text and data from scanned documents, including PDFs, and supports handwriting recognition via its 'forms' and 'tables' features. It uses machine learning to detect and extract printed text and handwritten content from form fields, making it the most appropriate choice for this use case.

Answer

Amazon Transcribe

Answer

Amazon Comprehend

Answer

Amazon Rekognition

Question 12

A company wants to build a customer service chatbot that answers questions about their internal policy documents. The documents are updated monthly, and the team cannot afford to retrain a model each time. Which approach is MOST appropriate?

Accepted Answer

Use Retrieval-Augmented Generation (RAG) with the policy documents indexed in a vector store. Retrieval-Augmented Generation (RAG) is the most appropriate approach because it allows the chatbot to answer questions based on the latest policy documents without retraining the underlying language model. By indexing the documents in a vector store and retrieving relevant chunks at query time, RAG provides up-to-date, grounded responses while keeping the base LLM static, which is both cost-effective and scalable for monthly document updates.

Answer

Fine-tune a base LLM on the policy documents monthly

Answer

Use a larger foundation model with a longer context window and paste all documents into each prompt

Answer

Train a custom model from scratch on the policy documents each month

Question 13

Which metric is commonly used to evaluate the quality of a text summarization model by comparing the generated summary with a reference summary, measuring the overlap of n-grams?

Accepted Answer

ROUGE. ROUGE (Recall-Oriented Understudy for Gisting Evaluation) is the standard metric for evaluating text summarization by measuring the overlap of n-grams (e.g., ROUGE-1, ROUGE-2, ROUGE-L) between a generated summary and a reference summary. It focuses on recall, capturing how much of the reference content is preserved in the generated summary, making it ideal for summarization tasks.

Answer

Perplexity

Answer

BLEU

Answer

BERTScore

Question 14

A team is deploying a foundation model on Amazon Bedrock for a customer-facing assistant. They must reduce hallucinations and keep answers grounded in approved company content while controlling inference cost. Which TWO approaches should the team implement? (Choose two.)

Accepted Answer

Configure Amazon Bedrock Guardrails contextual grounding and relevance checks to filter ungrounded responses.. Grounding answers in approved content through Bedrock knowledge bases supplies factual context, while Guardrails contextual grounding and relevance checks verify that generated responses are supported by that context. Together they reduce hallucination and keep outputs aligned with company material. Raising temperature, removing instructions, or simply choosing a larger model does not provide grounding and can increase cost or fabrication risk.

Answer

Remove all system instructions so the model relies solely on its pretrained knowledge.

Answer

Switch to the largest available foundation model to guarantee factual accuracy.

Answer

Increase the model temperature to encourage more creative and varied responses.

Question 15

A retail company is using Amazon Bedrock to generate personalized product recommendations in real time. The model sometimes produces recommendations that include products the company no longer sells. The company wants to ensure that only in-stock products are recommended. The product catalog is stored in an Amazon DynamoDB table that is updated continuously. Which solution should the company implement to meet this requirement with minimal latency?

Accepted Answer

Use Amazon Bedrock agents with an action group that queries DynamoDB for in-stock products, and filter the model's output accordingly.. To ensure recommendations only include in-stock products with minimal latency, the company should use Amazon Bedrock agents with an action group that queries DynamoDB. The agent can validate inventory in real time and filter the model's output. DynamoDB offers low-latency reads, making this suitable for real-time recommendations. Fine-tuning, top-p adjustments, or knowledge bases do not provide the necessary real-time inventory validation.

Answer

Fine-tune the foundation model daily with the latest product catalog from DynamoDB.

Answer

Increase the model's top-p parameter to include more diverse product recommendations.

Answer

Store the product catalog in an Amazon Bedrock knowledge base and rely on the model to retrieve only in-stock items.

Question 16

A financial services firm is deploying a generative AI assistant that answers employee questions about internal policies. Compliance requires that every response cite the exact policy document and section used. The assistant currently relies only on the foundation model's pretrained knowledge and frequently invents policy details. Which technique should the team implement to ground responses in the firm's own documents and produce citations?

Accepted Answer

Implement Retrieval Augmented Generation by embedding the policy documents in a vector store and retrieving relevant passages at query time.. Grounding a model in proprietary content requires supplying that content at inference time rather than relying on pretrained weights. Retrieval Augmented Generation retrieves the most relevant document chunks and places them in the prompt, so the model's answer is conditioned on real policy text and can reference the source document and section. Sampling or length adjustments do not add knowledge.

Answer

Increase the model's temperature setting so it explores more of its pretrained knowledge about corporate policy.

Answer

Lower the maximum token limit so the assistant gives shorter answers that are less likely to contain errors.

Answer

Fine-tune the foundation model on a large corpus of public financial regulations.

Question 17

A company runs a Retrieval Augmented Generation (RAG) application on Amazon Bedrock. The application uses an Amazon OpenSearch Serverless vector index to store internal HR documents. The security team must ensure that the vector index is encrypted at rest with a customer-managed AWS KMS key so that they can control key rotation and revoke access independently. Which configuration should they implement?

Accepted Answer

Create a customer managed KMS key and associate it with the OpenSearch Serverless collection at creation time using the encryption policy.. A customer managed KMS key must be specified in the OpenSearch Serverless encryption policy when the collection is created, because the encryption key cannot be changed after creation. This gives the security team control over rotation and the ability to revoke access by disabling the key. Encrypting source documents or using private connectivity does not encrypt the vector index at rest with a customer managed key.

Answer

Configure the OpenSearch Serverless collection with an AWS owned key and enable automatic key rotation in AWS KMS.

Answer

Enable Amazon S3 default encryption with SSE-KMS on the bucket that stores the source documents, then rebuild the vector index.

Answer

Use AWS PrivateLink to connect to the OpenSearch Serverless collection and rely on TLS in transit for data protection.

Question 18

A financial services company has built an internal assistant on Amazon Bedrock using Anthropic Claude 3 Sonnet. Employees ask questions that require retrieving the latest internal policy documents, which are updated frequently and stored in Amazon S3. The company wants the assistant to answer with accurate, up-to-date citations without retraining the model. Which approach should they implement?

Accepted Answer

Use Retrieval Augmented Generation (RAG) by creating a knowledge base in Amazon Bedrock that indexes the S3 documents.. Grounding a foundation model in frequently changing internal documents is best achieved with Retrieval Augmented Generation. An Amazon Bedrock knowledge base ingests the S3 content, chunks and embeds it, and retrieves relevant passages at inference time so the model can answer with current, citable information. This avoids retraining and keeps responses accurate as policies evolve.

Answer

Increase the temperature parameter to encourage the model to generate more detailed and specific policy answers.

Answer

Enable model invocation logging in Amazon Bedrock to capture requests and responses for later review.

Answer

Fine-tune the Claude 3 Sonnet model on the policy documents using Amazon Bedrock custom models.

Question 19

A media company uses Amazon Bedrock to generate short product descriptions. They notice that outputs vary in tone and sometimes include unwanted promotional claims. They want consistent, brand-aligned results while keeping the same foundation model. Which action best addresses this requirement?

Accepted Answer

Use a prompt template that specifies brand voice, required structure, and forbidden claims, and reuse it for every generation request.. Consistent brand-aligned generation is achieved by controlling the instructions the model receives. A reusable prompt template that defines voice, structure, and forbidden claims gives the model explicit guidance on every request, producing more uniform output. Guardrails, sampling parameters, and larger models can support quality but do not by themselves encode brand style and claim restrictions.

Answer

Switch to a larger foundation model with more parameters to improve output quality and consistency.

Answer

Create a guardrail in Amazon Bedrock with a denied topics policy and a word filter for prohibited promotional terms.

Answer

Lower the top-p value to 0.1 so the model only considers the most probable tokens.

Question 20

A company wants its Amazon Bedrock application to always answer in a formal tone, never discuss competitors, and never reveal internal project codenames. The controls must apply consistently to every request and response without changing the underlying model. Which Amazon Bedrock capability should the company configure?

Accepted Answer

Amazon Bedrock Guardrails. Amazon Bedrock Guardrails enforce content policies such as denied topics, word filters, and sensitive information detection on inputs and outputs independently of the model, so the same rules apply across every request. Sampling parameters, throughput reservations, and context window size influence generation behavior or capacity but cannot block topics or enforce tone consistently.

Answer

A higher temperature setting

Answer

Provisioned Throughput

Answer

A larger context window