AIF-C01 Practice Test 34 — 15 Questions

Question 1

A developer is using the Amazon Bedrock Converse API to build a chat application. They want the model to maintain context across multiple turns. Which parameter should they set to ensure the conversation history is included?

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

Accepted Answer

messages. The Converse API's `messages` parameter accepts an array of message objects, each with a `role` (user or assistant) and `content`. By including previous turns in this array, the model receives the full conversation history as context, enabling coherent multi-turn dialogue. Without it, each API call is stateless and the model has no memory of prior exchanges.

Answer

inferenceConfig

Answer

additionalModelRequestFields

Answer

system

Question 2

A company has a large dataset of customer support emails labeled with issue categories. They need to classify new emails automatically. Which algorithm is BEST suited for this task?

Accepted Answer

Logistic regression. Logistic regression is the best choice because it is a supervised learning algorithm specifically designed for binary or multi-class classification tasks. Given labeled emails with issue categories, logistic regression models the probability that a new email belongs to each category using a logistic (sigmoid) function, making it ideal for this classification problem.

Answer

Linear regression

Answer

K-means clustering

Answer

Gradient boosting

Question 3

During a binary classification project, the team wants to optimize for correctly identifying positive cases even if it means more false positives. Which metric should they maximize?

Accepted Answer

Recall. Recall (also known as sensitivity or true positive rate) measures the proportion of actual positive cases that are correctly identified. By maximizing recall, the model minimizes false negatives, which aligns with the goal of catching as many true positives as possible, even at the cost of increasing false positives.

Answer

Precision

Answer

AUC-ROC

Answer

F1 score

Question 4

A company uses Amazon Bedrock to generate content and wants to prevent the model from producing harmful or biased responses. Which AWS service should they configure to enforce content safety policies?

Accepted Answer

Amazon Bedrock Guardrails. Amazon Bedrock Guardrails provides content filtering, PII redaction, topic restrictions, and other safety controls. It is purpose-built for content safety in Bedrock.

Answer

Amazon Comprehend for toxicity detection

Answer

Amazon SageMaker Clarify

Answer

AWS WAF to filter model responses

Question 5

A team is building a deep learning model for image segmentation using Amazon SageMaker. They want to improve the model's generalization and reduce overfitting. Which THREE techniques should they apply? (Select THREE.)

Accepted Answer

L2 regularization on weights. L2 regularization on weights (A) is correct because adding a penalty proportional to the squared magnitude of weights (controlled by the regularization hyperparameter) constrains model complexity and directly reduces overfitting, improving generalization. Data augmentation (C) is correct because applying random crops, flips, and similar transformations synthetically expands the training set with label-preserving variations, which teaches the segmentation model invariance and reduces overfitting. Dropout regularization (E) is correct because randomly deactivating a fraction of neurons during training prevents co-adaptation of features and acts as an ensemble-like regularizer that improves generalization. Increasing the number of layers (B) is not appropriate here because deeper networks add capacity and typically worsen overfitting when data is limited. Using a smaller batch size (D) is not a regularization technique for reducing overfitting; it mainly affects gradient noise and training dynamics, not the model's generalization directly.

Answer

Increase the number of layers in the network

Answer

Use a smaller batch size

Question 6

A developer is creating a generative AI application using Amazon Bedrock and needs to ensure that responses do not include toxic or harmful content. Which feature should be enabled?

Accepted Answer

Amazon Bedrock Guardrails.. Amazon Bedrock Guardrails is the correct feature because it is specifically designed to enforce content policies, filter toxic or harmful content, and block undesirable topics in generative AI responses. It provides configurable thresholds for hate, insults, sexual content, violence, and other harmful categories, ensuring compliance with safety requirements without modifying the underlying model.

Answer

Amazon CloudWatch Logs for prompt logging.

Answer

Amazon Virtual Private Cloud (VPC) for network isolation.

Answer

AWS Identity and Access Management (IAM) policies.

Question 7

A data scientist has trained a random forest model that achieves 92% accuracy on the training set but only 75% on the test set. The dataset has 1000 samples and 20 features. Which THREE actions could help improve the model's generalization? (Select THREE.)

Accepted Answer

Increase the number of trees in the forest. The scenario describes overfitting: a large gap between training accuracy (92%) and test accuracy (75%). Option A (increase the number of trees in the forest) is correct because adding trees to a random forest averages more decorrelated trees, reducing variance without increasing overfitting, which typically improves test-set generalization. Option C (reduce the maximum depth of each tree) is correct because shallower trees have lower variance and cannot memorize training noise as easily, directly countering the overfitting. Option E (increase the minimum number of samples required to split an internal node) is correct because requiring more samples per split (e.g., raising min_samples_split) forces splits to be based on more evidence, producing simpler, more generalizable trees. Option B (increase the maximum depth) is wrong because deeper trees increase variance and worsen overfitting. Option D (decrease the number of trees) is wrong because fewer trees reduce the averaging benefit and can increase variance, hurting generalization.

Answer

Increase the maximum depth of each tree

Answer

Decrease the number of trees in the forest

Question 8

A company needs to convert live customer support calls into text for analysis and search. Which TWO AWS services should be used together? (Select TWO.)

Accepted Answer

Amazon Comprehend. Amazon Transcribe [CORRECT] is the right service for converting live audio from customer support calls into text, since it provides automatic speech recognition (ASR) that supports real-time streaming transcription as well as batch transcription of audio files. Amazon Comprehend [CORRECT] is then used on that transcribed text to perform natural language processing tasks such as entity recognition, key phrase extraction, sentiment analysis, and topic modeling, which directly supports the requirement to analyze and search the call content. Together they form the standard AWS pipeline for speech-to-text followed by text analytics. Amazon Rekognition is for image and video analysis, not audio transcription, so it does not fit. Amazon Polly performs text-to-speech synthesis (the reverse direction), and Amazon Translate performs language translation, neither of which is needed to convert calls to searchable, analyzable text.

Answer

Amazon Rekognition

Answer

Amazon Polly

Answer

Amazon Translate

Question 9

A security team needs to detect anomalies in AWS CloudTrail logs to identify potential unauthorized access. They want to use machine learning without manually labeling data or training custom models. Which AWS service should they use?

Accepted Answer

Amazon GuardDuty. Amazon GuardDuty is a threat detection service that uses machine learning and anomaly detection to continuously monitor AWS CloudTrail logs, VPC Flow Logs, and DNS logs for unauthorized access and malicious activity. It requires no manual labeling of data or custom model training, as it leverages pre-built ML models and integrated threat intelligence to identify suspicious behavior.

Answer

Amazon Macie

Answer

Amazon SageMaker

Answer

Amazon Detective

Question 10

A data scientist is fine-tuning a large language model on a custom dataset of legal documents. The dataset contains 100,000 documents, each with an average length of 5,000 tokens. The model has a context window of 8,192 tokens. What is the MOST important consideration for preparing the data for fine-tuning?

Accepted Answer

Split documents into chunks of fewer than 8,192 tokens with overlap. The model's context window is 8,192 tokens, but the average document length is 5,000 tokens, meaning many documents will exceed this limit. Splitting documents into chunks of fewer than 8,192 tokens with overlap ensures that no input exceeds the context window, preserves contextual continuity across chunk boundaries, and prevents loss of information that would occur with simple truncation.

Answer

Fine-tune the model on the full documents without modification

Answer

Convert documents into embeddings before fine-tuning

Answer

Truncate each document to 8,192 tokens

Question 11

A company is building a real-time document analysis tool using Amazon Bedrock. Their documents average 15,000 tokens each. Users submit a document and ask a single question about it. The team wants to minimize latency while maintaining answer quality. Which approach is MOST suitable?

Accepted Answer

Use Amazon Bedrock's Converse API with a prompt that includes the full document text and the user's question in a single invocation.. Using Amazon Bedrock's Converse API with a prompt that includes the full document text and the user's question in a single invocation minimizes latency by avoiding the overhead of multiple API calls or retrieval steps. This approach maintains answer quality because the model has access to the entire document context (up to 15,000 tokens) in one pass, which is well within the context window of models like Claude 3 or Llama 2, ensuring accurate responses without the need for chunking or external retrieval.

Answer

Use Amazon Bedrock's Knowledge Base (RAG) to retrieve relevant chunks from the document and then answer the question.

Answer

Split the document into chunks of 1,000 tokens each, process each chunk with the model separately, and aggregate results.

Answer

Use a fine-tuned model that has been trained on similar documents to avoid context processing.

Question 12

A company wants to automatically categorize customer feedback emails into topics such as 'billing', 'technical support', and 'general inquiry'. Which AWS service is MOST appropriate?

Accepted Answer

Amazon Comprehend. Amazon Comprehend is a natural language processing (NLP) service that uses machine learning to extract insights from text, including custom classification. It can be trained to automatically categorize customer feedback emails into predefined topics like 'billing', 'technical support', and 'general inquiry' by analyzing the text content and assigning it to the most relevant category.

Answer

Amazon Transcribe

Answer

Amazon Polly

Answer

Amazon Rekognition

Question 13

A retail company wants to forecast product demand at the SKU level for the next 12 weeks. Which AWS service is purpose-built for this task?

Accepted Answer

Amazon Forecast. Amazon Forecast is a fully managed service purpose-built for time-series forecasting, using machine learning to analyze historical data and predict future demand. It is specifically designed for tasks like SKU-level product demand forecasting over a defined time horizon, such as 12 weeks, without requiring deep ML expertise.

Answer

Amazon Personalize

Answer

Amazon SageMaker

Answer

Amazon Kendra

Question 14

Which type of machine learning is used when a model learns to play a game by receiving rewards or penalties for its actions?

Accepted Answer

Reinforcement learning. Reinforcement learning is correct because it involves an agent learning to make decisions by interacting with an environment, receiving rewards for desirable actions and penalties for undesirable ones. This trial-and-error approach, often modeled using Markov Decision Processes (MDPs), is specifically designed for sequential decision-making tasks like game playing, where the agent must maximize cumulative reward over time.

Answer

Semi-supervised learning

Answer

Supervised learning

Answer

Unsupervised learning

Question 15

A developer is selecting a foundation model on Amazon Bedrock for a real-time text summarization application. Which THREE factors should they consider when choosing the model? (Choose three.)

Accepted Answer

Latency and throughput requirements. Option B is correct because a real-time text summarization application demands low latency and high throughput, so the developer must evaluate each Bedrock foundation model's inference speed and tokens-per-second capacity to meet interactive response times. Option D is correct because Bedrock charges based on input and output tokens, so the cost per token directly affects the operating budget of a high-volume summarization workload. Option E is correct because the model must support the required output modality—text (and possibly code)—for summarization; models limited to other modalities such as embeddings or image generation would not satisfy the use case. Option A is incorrect because a model's logo color is purely cosmetic branding and has no bearing on technical or business fit. Option C is incorrect because image generation is irrelevant to text summarization, and selecting a model for that capability would waste cost and latency on unneeded functionality.

Answer

The color of the model's logo

Answer

Model's ability to generate images