AIF-C01 Practice Exam 51 — 20 Questions

Question 1

Refer to the exhibit. A data scientist runs an Amazon SageMaker Clarify bias analysis on a binary classifier. The pre-training ClassImbalance is 1.5 and the post-training DPPL is 0.15. What should the data scientist conclude?

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

Accepted Answer

The data has a mild class imbalance, but the model shows a noticeable bias in predictions.. The pre-training ClassImbalance metric of 1.5 indicates a mild class imbalance (values close to 1.0 indicate balance, while values significantly above 1.0 indicate imbalance). The post-training DPPL (Difference in Positive Proportions in Labels) metric of 0.15 exceeds the commonly accepted fairness threshold of 0.10, indicating a noticeable bias in the model's predictions. Therefore, the data has a mild imbalance, but the model exhibits a bias that warrants further investigation.

Answer

The data is highly imbalanced and the model is unbiased.

Answer

The pre-training metric indicates a fairness issue, but the post-training metric is acceptable.

Answer

The data is perfectly balanced and the model is fair.

Question 2

A data scientist is evaluating a regression model and wants to understand its prediction errors. Which TWO metrics should they use? (Select TWO.)

Accepted Answer

Mean Absolute Error (MAE). Mean Absolute Error (MAE) is correct because it directly quantifies regression prediction error as the average of the absolute differences between predicted and actual values, expressed in the same units as the target variable. Root Mean Squared Error (RMSE) is also correct because it measures regression error as the square root of the mean of squared differences, penalizing larger errors more heavily and remaining in the target's units. Both MAE and RMSE are standard regression error metrics that summarize how far predictions deviate from actual continuous values. F1 score is incorrect because it is a classification metric combining precision and recall, not applicable to continuous regression errors. Accuracy is incorrect because it measures the proportion of correct class predictions in classification. Precision is incorrect because it is a classification metric for the proportion of positive predictions that are truly positive, not a regression error measure.

Answer

F1 score

Answer

Accuracy

Answer

Precision

Question 3

A retail company wants to generate product descriptions from catalog data. The data includes structured attributes (e.g., price, brand) and unstructured reviews. The team needs to ensure factual accuracy. Which approach is most appropriate?

Accepted Answer

Implement Retrieval-Augmented Generation (RAG) with a knowledge base. Retrieval-Augmented Generation (RAG) retrieves relevant documents (product attributes, reviews) and provides them as context to the model, reducing hallucinations and grounding responses in facts.

Answer

Use prompt engineering with few-shot examples

Answer

Fine-tune a foundation model on the entire product catalog

Answer

Deploy a larger foundation model with more parameters

Question 4

A company is building a chatbot using Amazon Bedrock and wants to ensure that the model generates responses consistent with its brand voice. Which technique should be used to provide the model with examples of desired responses without fine-tuning the model?

Accepted Answer

Include few-shot examples in the system prompt to demonstrate the desired tone.. Few-shot prompting allows you to provide the model with examples of desired responses directly in the system prompt, guiding the model's tone and style without modifying its underlying weights. This technique is ideal for brand voice consistency when fine-tuning is not an option, as it leverages in-context learning to influence output behavior.

Answer

Fine-tune the model on a dataset of brand-compliant conversations.

Answer

Use prompt chaining to break down the conversation into multiple steps.

Answer

Implement a Retrieval Augmented Generation (RAG) system with brand documents.

Question 5

A developer is using Bedrock Agents to create an order management assistant. The agent needs to check inventory and process payments. Which TWO components are required to enable these capabilities?

Accepted Answer

Lambda functions to execute the API calls. Option B is correct because Bedrock Agents use AWS Lambda functions as the compute layer that actually executes the business logic and outbound API calls — such as querying an inventory system or invoking a payment gateway — when the agent decides an action is needed. Option D is correct because action groups define the agent's callable capabilities, and each action group is described by an OpenAPI schema (API schema) that tells the agent which operations exist, their parameters, and how to invoke the backing Lambda, which is exactly what enables inventory checks and payment processing. Option A is not required here because Knowledge Bases are for retrieval-augmented generation over documents (e.g., product info), not for performing transactional API operations. Option C is not required because a vector store is used for embeddings/semantic search in Knowledge Bases, not for caching transaction data. Option E is not required because Guardrails apply content and safety filtering to inputs/outputs, which is unrelated to enabling inventory and payment actions.

Answer

Knowledge Bases for product information

Answer

A vector store to cache transaction data

Answer

Guardrails to filter sensitive data

Question 6

A financial services company is subject to strict regulatory requirements. They plan to use generative AI to summarize customer interaction logs. Which combination of AWS services and configurations best ensures compliance while maintaining accuracy?

Accepted Answer

Use Amazon Bedrock with a private VPC endpoint, AWS KMS encryption, and content filtering.. It combines a private VPC endpoint to keep all traffic within the AWS network (avoiding public internet exposure), AWS KMS encryption for data at rest and in transit, and content filtering to block sensitive or non-compliant outputs. This architecture meets strict regulatory requirements for data privacy and security while using Amazon Bedrock's managed foundation models for accurate summarization.

Answer

Deploy an open-source model on Amazon Bedrock in a local on-premises server.

Answer

Use Amazon Bedrock with a foundation model and public internet access without encryption.

Answer

Use Amazon SageMaker to host a fine-tuned model with a public API key.

Question 7

A team needs to identify customer segments based on purchasing behavior without predefined categories. Which algorithm should they use?

Accepted Answer

K-means clustering. K-means clustering is an unsupervised learning algorithm that groups data points into clusters based on feature similarity without requiring predefined labels or categories. Since the team needs to identify customer segments solely from purchasing behavior data, K-means is the appropriate choice as it discovers natural groupings in the data.

Answer

Linear regression

Answer

Decision tree

Answer

Logistic regression

Question 8

What is a foundation model?

Accepted Answer

A model trained on diverse data that can be adapted to many tasks. A foundation model is a large-scale AI model trained on vast, diverse datasets (e.g., text, images, code) using self-supervised learning, enabling it to be adapted to a wide range of downstream tasks through fine-tuning or few-shot learning. Option C correctly captures this core property of broad adaptability, which distinguishes foundation models from task-specific models. For example, GPT-4 and Claude are foundation models that can handle translation, summarization, and coding without being retrained from scratch.

Answer

A model that only works with tabular data

Answer

A model that requires no additional tuning for new tasks

Answer

A model that is specifically trained for one task, like image classification

Question 9

A healthcare company is deploying a conversational AI using a foundation model on Amazon Bedrock for patient triage. The application must minimize hallucinations and ensure factual accuracy. Which combination of techniques should the team implement?

Accepted Answer

Implement Retrieval-Augmented Generation (RAG) using a knowledge base on Amazon Bedrock and a system prompt demanding factual responses.. Retrieval-Augmented Generation (RAG) with a knowledge base on Amazon Bedrock grounds the model's responses in authoritative, up-to-date documents, dramatically reducing hallucinations. Pairing RAG with a system prompt that demands factual, sourced answers further constrains the model. This combination directly addresses the requirement to minimize hallucinations and ensure factual accuracy for patient triage.

Answer

Fine-tune the model on a large dataset of medical transcripts and deploy with default parameters.

Answer

Use reinforcement learning from human feedback (RLHF) on the deployed model.

Answer

Set the maxTokens to a low value to force shorter, more focused answers.

Question 10

A company deployed a question-answering system using Amazon Bedrock with a knowledge base (RAG). Users report that the model often hallucinates facts not in the knowledge base. What is the most effective way to reduce hallucinations?

Accepted Answer

Improve the relevance of retrieved documents by refining the retrieval strategy. Hallucinations in RAG systems often stem from the model receiving irrelevant or low-quality retrieved documents, which forces it to rely on its parametric knowledge rather than the provided context. By refining the retrieval strategy—such as improving embedding quality, adjusting chunk overlap, or using hybrid search—the system ensures the foundation model has the most relevant information to ground its answers, directly reducing the likelihood of fabricating facts.

Answer

Reduce the maximum context length to limit model input

Answer

Fine-tune the foundation model on a large general corpus

Answer

Increase the chunk size of documents in the knowledge base

Question 11

A company uses Amazon Bedrock to build an AI assistant. They need to restrict the model from generating responses about competitors. Which Bedrock feature should they configure?

Accepted Answer

Denied topics. Bedrock Guardrails allows you to define topic restrictions that block certain topics from being discussed. Other options are for content filtering, PII redaction, or grounding.

Answer

Word filters

Answer

PII redaction

Answer

Content filters

Question 12

A company is deploying a customer-facing chatbot using Amazon Bedrock. They want to reduce the risk of generating biased or harmful responses. Which TWO measures should they implement? (Choose 2.)

Accepted Answer

Implement a human-in-the-loop review for sensitive replies. Option A is correct because implementing a human-in-the-loop review for sensitive replies adds a manual verification layer that can catch biased or harmful outputs before they reach customers, which is a recommended responsible-AI practice for customer-facing generative applications. Option C is correct because Amazon Bedrock Guardrails lets you configure content filters (for hate, insults, sexual, violence, misconduct), denied topics, word filters, and contextual grounding checks that automatically block or mask harmful or biased responses at inference time. Option B is not appropriate because training exclusively on historical customer conversations can perpetuate existing biases and does not by itself mitigate harmful output. Option D is incorrect because setting temperature to 1.5 increases randomness and creativity, making outputs less predictable and potentially more harmful. Option E is incorrect because disabling logging reduces observability and auditability, which undermines monitoring and continuous improvement of safety controls.

Answer

Train the model exclusively on historical customer conversations

Answer

Set the temperature parameter to 1.5

Answer

Disable logging to improve performance

Question 13

A machine learning engineer notices that the training loss decreases steadily, but the validation loss starts increasing after a few epochs. Which of the following is the MOST likely cause?

Accepted Answer

Overfitting. The scenario describes training loss decreasing while validation loss increases after a few epochs, which is the classic signature of overfitting. The model is memorizing the training data (including noise) rather than learning generalizable patterns, causing it to perform poorly on unseen validation data.

Answer

Learning rate too low

Answer

Underfitting

Answer

Data leakage from validation set into training set

Question 14

A company wants to detect sensitive data such as PII in their training datasets stored in S3 before using them for model training. Which AWS service should they use?

Accepted Answer

Amazon Macie. Amazon Macie uses machine learning to discover and protect sensitive data in S3, including PII.

Answer

Amazon Inspector

Answer

AWS Shield

Answer

Amazon GuardDuty

Question 15

A financial services company wants to generate personalized investment recommendations using a large language model via Amazon Bedrock. They have customer data that includes risk tolerance, portfolio holdings, and financial goals. The company is highly concerned about data privacy and must avoid exposing sensitive personally identifiable information (PII) to the model. They plan to use a foundation model to generate recommendations based on customer profiles. What is the best approach to protect customer privacy while still enabling personalization?

Accepted Answer

Preprocess the customer data to replace sensitive fields with placeholders, then use the processed data in the prompt.. Preprocessing customer data to replace sensitive fields with placeholders (e.g., using synthetic IDs) allows the model to generate personalized recommendations without accessing real PII. This minimizes risk. Option A is incorrect because fine-tuning on a large dataset of generic recommendations does not produce personalized outputs for individual customers. Option B is incorrect because prompt engineering instructions are not a robust privacy control and cannot reliably prevent PII exposure. Option D is incorrect because including customer data directly in the prompt and relying on the model to anonymize it is unreliable and may still leak PII.

Answer

Fine-tune the model on a large dataset of investment recommendations without any customer-specific data.

Answer

Use prompt engineering to instruct the model to disregard any personally identifiable information.

Answer

Include the customer data directly in the prompt and rely on the model to anonymize it.

Question 16

A startup is building an AI-powered code assistant using a large language model (LLM). They want to ensure the model generates syntactically correct code and avoids security vulnerabilities. Which technique should they prioritize?

Accepted Answer

Augment prompts with few-shot examples of secure coding practices and unit tests. Few-shot prompting with examples of secure coding practices and unit tests directly shapes the model's output distribution toward syntactically valid, security-conscious code by conditioning on concrete in-context demonstrations. This is a prompt-engineering technique that requires no retraining and can be updated instantly as new vulnerability patterns emerge. It simultaneously addresses both stated goals: syntactic correctness (via code examples) and security (via secure-coding exemplars).

Answer

Deploy the model with max tokens set to 4096

Answer

Fine-tune the model on a large corpus of open-source code

Answer

Use chain-of-thought prompting to explain reasoning before code generation

Question 17

A machine learning engineer is training a neural network and wants to prevent overfitting. Which technique should they apply?

Accepted Answer

Dropout. Dropout is a regularization technique that randomly drops a fraction of neurons during training, which prevents the network from becoming overly reliant on any single neuron and reduces co-adaptation. This forces the model to learn more robust features, effectively reducing overfitting by acting as an ensemble of sub-networks.

Answer

Gradient descent

Answer

Backpropagation

Answer

Boosting

Question 18

A bank is using Amazon Bedrock to summarize customer support transcripts. The summaries often contain factual inaccuracies (hallucinations). Which approach is most effective for reducing hallucinations?

Accepted Answer

Implement RAG by grounding summarization on retrieved transcripts. Retrieval-Augmented Generation (RAG) grounds the model's output on actual retrieved chunks of the customer support transcripts, providing factual context that reduces the likelihood of hallucination. By retrieving relevant transcript segments and feeding them as context to the LLM, the model generates summaries based on verified source material rather than relying solely on its parametric knowledge, which is the primary cause of factual inaccuracies.

Answer

Decrease the top-p to 0.1

Answer

Increase the model's temperature to make outputs more diverse

Answer

Fine-tune a smaller model on a large dataset of transcripts

Question 19

Which AWS service can be used to create a personalized recommendation engine for an e-commerce website without requiring prior machine learning experience?

Accepted Answer

Amazon Personalize. Amazon Personalize is a fully managed machine learning service designed specifically to build personalized recommendation engines (e.g., product recommendations, personalized content) without requiring prior ML expertise. It uses the same technology that powers Amazon.com's recommendations, providing pre-built algorithms and automatic model training, tuning, and deployment via a simple API.

Answer

Amazon SageMaker

Answer

Amazon Rekognition

Answer

Amazon Forecast

Question 20

A healthcare startup is building a patient inquiry system using Amazon Bedrock. They must ensure the model does not generate responses containing medical advice or unverified treatment suggestions. The compliance team also requires that no personally identifiable information (PII) is output. Which Bedrock feature should the startup configure to meet both requirements?

Accepted Answer

Bedrock Guardrails with topic denial and PII detection. Bedrock Guardrails provides content filtering, topic denial (to block medical advice), and PII detection/redaction. Topic denial can block entire categories of unwanted responses, and PII detection prevents leakage of personal data.

Answer

Bedrock Knowledge Bases with a custom chunking strategy

Answer

Amazon Comprehend for PII detection and a separate content moderation service

Answer

Bedrock Agents with a Lambda function that validates responses