AIF-C01 Practice Exam 54 — 20 Questions

Question 1

A financial services company is using Amazon Bedrock to generate personalized investment advice. The compliance team requires that the model's responses do not contain any personally identifiable information (PII) such as account numbers or social security numbers, and that all PII is automatically masked. Which AWS service or feature should the company use to meet this requirement?

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

Accepted Answer

Amazon Bedrock Guardrails. Amazon Bedrock Guardrails provides configurable safeguards that can detect and mask sensitive information like PII in both prompts and responses. By defining a sensitive information filter, the company can ensure that account numbers and social security numbers are automatically redacted from the model's output. This is the only option that directly addresses real-time content filtering and masking within the generative AI application.

Answer

Amazon Macie

Answer

AWS Identity and Access Management (IAM) policies

Answer

Amazon SageMaker Model Monitor

Question 2

A financial services company is using Amazon Bedrock to generate investment summaries. They must ensure that the model does not provide personalized financial advice, which is a regulatory requirement. Which AWS feature should they use to block the model from generating such advice?

Accepted Answer

Amazon Bedrock Guardrails with a denied topics policy.. To block the model from generating personalized financial advice, the company should use Amazon Bedrock Guardrails with a denied topics policy. This feature allows defining topics that the model must avoid, effectively preventing non-compliant responses. Other options like model evaluation or fine-tuning do not provide real-time blocking.

Answer

Amazon Bedrock model evaluation with a custom metric for regulatory compliance.

Answer

Amazon Bedrock Agents with an action group that checks for financial advice keywords.

Answer

Amazon Bedrock Provisioned Throughput with a custom model that has been fine-tuned to avoid financial advice.

Question 3

A developer is building a prototype that needs to generate product descriptions from a few keywords. The developer wants to use a foundation model on Amazon Bedrock but has no labeled training data and wants to avoid managing infrastructure. Which approach should the developer use?

Accepted Answer

Use the foundation model with a well-crafted prompt that includes the keywords and desired format.. For a prototype with no labeled data and a desire to avoid infrastructure management, using a foundation model on Amazon Bedrock with a well-designed prompt is ideal. Prompt engineering allows the model to generate product descriptions from keywords without any training. Fine-tuning, training from scratch, or self-managing an EC2 deployment all introduce unnecessary complexity and data requirements that the scenario does not support.

Answer

Deploy an open-source model on an Amazon EC2 instance and write custom inference code.

Answer

Train a custom model from scratch using Amazon SageMaker.

Answer

Fine-tune a foundation model with a small dataset of product descriptions.

Question 4

A company is deploying a foundation model on Amazon Bedrock to generate product descriptions. They want to ensure the model's output is factually consistent with the provided product specifications and avoids hallucinated features. Which TWO techniques should they use? (Choose two.)

Accepted Answer

Use retrieval-augmented generation (RAG) with Amazon Bedrock Knowledge Bases to ground the model in the product specifications.. To ensure factual consistency, the company should ground the model in the actual product specifications. Retrieval-augmented generation with Amazon Bedrock Knowledge Bases provides relevant context, while clear prompt instructions that restrict the model to given information further reduce hallucinations. Both techniques work together to keep output aligned with facts.

Answer

Set the temperature to a high value to encourage the model to explore more creative descriptions.

Answer

Fine-tune the model on a dataset of existing product descriptions to improve its writing style.

Answer

Use Amazon Bedrock Guardrails to define a denied topic for features not present in the specifications.

Question 5

A company is deploying a machine learning model on Amazon SageMaker. The compliance team requires that the model's predictions be explainable and that the company can provide documentation on how the model makes decisions. Which SageMaker feature should the company use to meet this requirement?

Accepted Answer

Amazon SageMaker Clarify. SageMaker Clarify is specifically designed to provide explainability for machine learning models. It generates feature attribution reports that show the contribution of each input feature to the model's predictions, which can be used to document and explain model decisions. This meets the compliance requirement for explainable AI.

Answer

Amazon SageMaker Autopilot

Answer

Amazon SageMaker Debugger

Answer

Amazon SageMaker Model Monitor

Question 6

A financial services company uses Amazon Bedrock to power an internal assistant that answers employee questions about HR policies. The company must ensure that the assistant never reveals sensitive employee data. The HR policy documents are stored in an Amazon S3 bucket and are updated frequently. The company wants the assistant to cite the exact policy document and section for each answer. Which solution meets these requirements with the LEAST operational overhead?

Accepted Answer

Create an Amazon Bedrock knowledge base backed by the S3 bucket, and use the RetrieveAndGenerate API with citations enabled.. The company needs accurate, cited answers from frequently updated HR documents with minimal operational overhead. Amazon Bedrock knowledge bases integrate with S3, automatically manage indexing and retrieval, and support citations via the RetrieveAndGenerate API. This managed solution avoids custom infrastructure and ensures answers are grounded in the latest policies. Fine-tuning, custom retrieval, or relying on model training data all fail to meet the citation and maintenance requirements efficiently.

Answer

Build a custom retrieval system using Amazon OpenSearch Service and write a Lambda function to call the model with retrieved passages.

Answer

Use the model's built-in knowledge by prompting it to answer from its training data, and add a guardrail to filter sensitive information.

Answer

Fine-tune a foundation model on the HR policy documents and deploy it with a guardrail that blocks sensitive data.

Question 7

A healthcare provider wants to predict which patients are likely to be readmitted within 30 days. The historical dataset has 12,000 admissions and includes age, diagnosis codes, length of stay, and prior admissions. The team has limited machine learning experience and needs an explainable model that shows which factors drove each prediction. Which AWS approach is most appropriate?

Accepted Answer

Use Amazon SageMaker Autopilot to train candidate models on the labeled dataset, then review the model explainability report to understand feature influence.. The task is supervised binary classification on tabular patient data, which SageMaker Autopilot handles by automatically exploring preprocessing, algorithms, and hyperparameters. Its model explainability report attributes predictions to input features, satisfying the need for transparency. Forecasting, recommendation, and conversational services do not perform per-patient clinical risk prediction with feature-level explanations.

Answer

Use Amazon Personalize to rank patients by readmission risk based on their historical interactions.

Answer

Use Amazon Forecast to predict readmission counts and rely on its built-in accuracy metrics for explanation.

Answer

Use Amazon Lex to build a conversational interface that asks clinicians to estimate readmission probability.

Question 8

A retail company runs a product-question answering feature on Amazon Bedrock. During peak hours, requests intermittently fail with a ThrottlingException even though average usage is well within quota. The team needs a solution that smooths bursty traffic, retries failed calls, and avoids overwhelming the model endpoint, with minimal application code changes. Which approach should they take?

Accepted Answer

Introduce client-side retries with exponential backoff and jitter around the Amazon Bedrock InvokeModel calls.. Bursty traffic that briefly exceeds per-account or per-model quotas produces throttling that is transient by nature. Retrying with exponential backoff and jitter lets the client back off and spread retries, converting failures into eventual successes without architectural change. Provisioned Throughput, queues, and larger token limits either cost more, add latency, or do not address retry behaviour, so backoff and jitter is the targeted remedy.

Answer

Switch the application to call the model through Amazon Bedrock Provisioned Throughput and remove all retry logic.

Answer

Place an Amazon SQS queue between the application and Amazon Bedrock, and have a worker invoke the model as messages arrive.

Answer

Increase the maximum token count in each request so fewer total requests are needed during peak periods.

Question 9

A financial services company runs a fraud-detection model on an Amazon SageMaker endpoint. Auditors require the company to prove that each prediction can be traced back to the specific model artifact, container image, and training data version used, so the company must be able to reproduce and explain any past decision. Which AWS capability should the company implement to meet this requirement?

Accepted Answer

Configure SageMaker Model Registry with model cards and record lineage metadata in Amazon SageMaker ML Lineage Tracking.. Auditors need a defensible chain from prediction back to artifact, image, and data version. SageMaker Model Registry stores versioned, approved model packages with metadata, and ML Lineage Tracking records the relationships among datasets, jobs, artifacts, and endpoints. Using them together creates the reproducible evidence trail, whereas monitoring, access logging, or cost tagging each answer a different question.

Answer

Apply resource tags to the SageMaker endpoint and generate a cost allocation report in AWS Cost Explorer.

Answer

Enable SageMaker Model Monitor with a data quality baseline on the production endpoint.

Answer

Turn on AWS CloudTrail data events for the Amazon S3 bucket that stores the training data.

Question 10

A healthcare startup is using Amazon Bedrock to build a patient education chatbot. The chatbot must generate responses that are empathetic, accurate, and compliant with medical privacy regulations. The startup wants to implement safeguards to prevent the model from generating harmful or inappropriate content. Which TWO actions should the startup take to meet these requirements? (Choose two.)

Accepted Answer

Use Amazon Comprehend Medical to detect and redact protected health information (PHI) in user inputs and model outputs.. To prevent harmful content and ensure medical privacy compliance, the startup should use Amazon Bedrock Guardrails to filter harmful categories and denied topics, and Amazon Comprehend Medical to detect and redact PHI. Guardrails provide runtime content moderation, while Comprehend Medical specializes in medical data privacy. Together they address both safety and regulatory requirements. Other options like low temperature or prompt engineering do not enforce safeguards, and model evaluation is not a real-time filter.

Answer

Enable Amazon Bedrock model evaluation to automatically monitor and block harmful outputs in real time.

Answer

Use a foundation model with a low temperature setting to reduce randomness.

Answer

Implement prompt engineering to instruct the model to be empathetic and avoid medical advice.

Question 11

A media company uses Amazon Bedrock to generate short video scripts. Writers report that outputs are inconsistent in tone and sometimes ignore the required scene structure. The team wants a reusable, versioned artifact that enforces the role, tone, and output format across many invocations without retraining the model, and they want to track changes over time. Which Amazon Bedrock feature should they use?

Accepted Answer

Amazon Bedrock Prompt Management with prompt versions. A reusable, versioned instruction artifact is exactly what Amazon Bedrock Prompt Management provides. Prompts can be authored, stored, and referenced by ARN, with versions that capture changes over time, so every invocation uses the same role, tone, and structure guidance without touching model weights. Evaluation jobs, Guardrails, and Provisioned Throughput address quality measurement, output filtering, and capacity respectively, not prompt standardization.

Answer

Provisioned Throughput on the foundation model

Answer

Amazon Bedrock Guardrails with a denied topics policy

Answer

Amazon Bedrock model evaluation jobs

Question 12

A software company is deploying a generative AI assistant on Amazon Bedrock. They need the assistant to include citations to source documents in its answers and to avoid answering when no supporting document is found. Which configuration should they use?

Accepted Answer

Configure an Amazon Bedrock knowledge base and enable citation generation, and instruct the model to answer only from retrieved context.. An Amazon Bedrock knowledge base with citation generation returns responses that reference the retrieved source chunks, and a prompt that restricts answers to retrieved context makes the assistant abstain when no supporting document exists. This combination meets both the attribution and the no-support behavior requirements.

Answer

Enable model invocation logging and parse the logs to extract source references.

Answer

Use a higher temperature setting so the model explores more possible answers, including citations.

Answer

Increase the model's maximum token count so it can include citations in its response.

Question 13

A hospital is deploying an Amazon Bedrock-powered assistant that answers staff questions about internal policies. The compliance team requires that the assistant refuse requests for individual patient diagnoses and that any response containing protected health information be masked before it is returned. Which TWO capabilities should the team configure in Amazon Bedrock Guardrails to meet these requirements? (Choose two.)

Accepted Answer

Configure a sensitive information policy that detects and masks protected health information in model responses.. Guardrails enforce content policy on both inputs and outputs. A denied topics policy blocks requests and responses about individual patient diagnosis, while a sensitive information policy detects and masks protected health information in responses. Together they satisfy the refusal and masking requirements. Logging, temperature settings, and Provisioned Throughput do not prevent or redact content at inference time.

Answer

Enable model invocation logging to capture all prompts and responses for later compliance review.

Answer

Create a Provisioned Throughput purchase for the model to guarantee response capacity for clinical staff.

Answer

Set the model temperature to zero so responses are deterministic and cannot include patient details.

Question 14

A hospital wants to detect pneumonia from chest X-ray images. Radiologists have already labeled thousands of past X-rays as either 'pneumonia' or 'no pneumonia'. The hospital wants a model that generalizes to new X-rays. Which type of machine learning task is this?

Accepted Answer

Supervised image classification. Labeled images with known diagnoses are the defining input for supervised classification. The model learns from pneumonia and no-pneumonia examples and predicts the class for new X-rays, which is precisely the generalization the hospital wants. Unsupervised, clustering, and reinforcement approaches either ignore the labels or lack the required structure.

Answer

Reinforcement learning

Answer

Unsupervised anomaly detection

Answer

Clustering

Question 15

A hospital wants to build a system that automatically assigns a specialty department (for example, Cardiology, Neurology, or Orthopedics) to each free-text patient referral note. The hospital has a large archive of past referral notes that were already labeled by clinicians with the correct department. Which type of machine learning problem does this scenario describe?

Accepted Answer

Supervised learning, because the model learns from labeled historical examples to predict a categorical outcome.. Because every historical referral note already has a clinician-assigned department, the training data is fully labeled and the target is one of several discrete categories. That combination defines supervised classification, in which a model learns the relationship between note text and department and then predicts the department for new notes.

Answer

Generative AI, because the model must create a new department name for every referral note it processes.

Answer

Unsupervised learning, because the model must discover hidden structure within the referral notes without guidance.

Answer

Reinforcement learning, because the model improves through rewards received after each department assignment.

Question 16

A retail company wants to forecast daily product demand for the next quarter. They have three years of historical sales data that includes seasonal spikes and promotional periods, and they want a fully managed AWS service that can automatically train and tune a forecasting model without writing deep learning code. Which AWS service best fits this requirement?

Accepted Answer

Amazon Forecast. The scenario calls for a managed service that learns from historical numeric sales data containing seasonality and promotions. Amazon Forecast is purpose-built for this, automatically training and tuning time-series models and exposing forecasts without custom deep learning code. The other services address computer vision, speech, and text analytics, none of which produce demand forecasts.

Answer

Amazon Comprehend

Answer

Amazon Polly

Answer

Amazon Rekognition

Question 17

A startup prototypes a support assistant using Amazon Bedrock and needs to compare how three different foundation models handle the same set of 200 support tickets. They want objective quality scores, including accuracy against reference answers and robustness, before committing to one model. Which approach should they use?

Accepted Answer

Create an Amazon Bedrock model evaluation job with the tickets and reference answers.. The startup needs standardized, repeatable quality metrics across multiple models. Amazon Bedrock model evaluation jobs accept a dataset with prompts and reference responses and compute metrics such as accuracy, robustness, and toxicity for each selected model, making results directly comparable. Guardrail block rates, informal human ratings, and latency measurements capture different properties and cannot substitute for objective quality scoring.

Answer

Increase Provisioned Throughput for each model and measure response latency.

Answer

Enable Guardrails for Amazon Bedrock on each model and compare block rates.

Answer

Send the tickets through each model and have engineers rate the replies informally.

Question 18

A fintech startup is preparing its first machine learning project to detect fraudulent card transactions. The team must decide which characteristics make a problem well suited to supervised learning. Which TWO characteristics indicate that supervised learning is appropriate? (Choose two.)

Accepted Answer

Historical transactions exist that were confirmed as fraudulent or legitimate by investigators.. Supervised learning needs labeled examples and a defined target to predict. Confirmed fraudulent or legitimate transactions supply those labels, and a discrete fraud-or-not decision supplies a suitable classification target. Unlabeled data, label-free exploration goals, and explainability preferences do not establish the labeled input-output pairing that supervised learning fundamentally depends on.

Answer

The team wants the algorithm to discover hidden structure without any predefined output.

Answer

The team prefers a model that explains each prediction with feature importance values.

Answer

The dataset contains millions of unlabeled transactions with no investigator feedback.

Question 19

A media company wants to build a generative AI assistant that drafts scripts and answers questions about its own style guide. The team has no machine learning engineers and wants to avoid managing GPU infrastructure or training any models. Which approach BEST describes how they should build this solution?

Accepted Answer

Use a foundation model through Amazon Bedrock and supply the style guide as context at inference time.. A pretrained foundation model delivered through a managed service removes the need to train models or operate GPU infrastructure, and the proprietary style guide can be injected as prompt context so outputs follow house rules. This combination satisfies the no-ML-team constraint while still producing grounded, task-specific generative output.

Answer

Build a rules-based template engine that reassembles stored sentences from previous scripts.

Answer

Deploy an Amazon SageMaker endpoint hosting an open-source model and manually patch the underlying EC2 instances.

Answer

Collect a labeled dataset and train a new transformer from scratch on Amazon SageMaker training jobs.

Question 20

A healthcare startup uses Amazon Bedrock to power a patient-facing chatbot. Compliance officers are concerned that prompts or responses could contain protected health information and want an automated control that detects and blocks such content in real time before it reaches the model or the user. Which Amazon Bedrock feature should the team configure?

Accepted Answer

Amazon Bedrock Guardrails with sensitive information filters enabled.. The requirement is inline detection and blocking of sensitive content in both prompts and responses. Amazon Bedrock Guardrails applies configurable policies during inference, and its sensitive information filters identify PII and can mask or block it. Storage scanning, private networking, and offline evaluation each miss the real-time enforcement point where the chatbot interacts with patients.

Answer

Amazon Macie sensitive data discovery jobs scoped to the Bedrock service-linked bucket.

Answer

AWS PrivateLink interface endpoints for the Amazon Bedrock runtime.

Answer

Amazon Bedrock model evaluation jobs configured with an automatic toxicity metric.