A law firm wants to use an Amazon Bedrock knowledge base as a retrieval-augmented generation (RAG) solution. The knowledge base contains years of regulatory requirements. The law firm needs a solution to research regulatory requirements that were valid during a given time period. Many documents ingested into the knowledge base do not contain any dates in their content. Which solution will minimize operational overhead to meet these requirements?

D. Ingest metadata into the knowledge base and retrieve content only from documents tagged with specific dates

A. Build and deploy an AWS Lambda function to request specific document content from the knowledge base

B. Build and deploy containers on Amazon Elastic Container Service (ECS) to request specific document content from the knowledge base

C. Build and deploy containers on Amazon Elastic Kubernetes Service (EKS) to request specific document content from the knowledge base

D. Ingest metadata into the knowledge base and retrieve content only from documents tagged with specific dates

A data scientist wants to use a diffusion model to generate images from text prompts. The generated images are often blurry and contain unwanted noise. Which prompting technique can help remove unwanted noise?

D. Single-shot prompting with context

A company wants to build an internal chatbot application using Amazon Bedrock to answer employee questions based on internal documentation. The company aims to minimize application development costs. Which solution will minimize operational overhead to meet these requirements?

C. Create a knowledge base and use retrieval-augmented generation (RAG)

A. Use fine-tuning to adapt an LLM with the company's data

B. Use continual pre-training to adapt an LLM with the company's data

C. Create a knowledge base and use retrieval-augmented generation (RAG)

D. Train an LLM from scratch using only the company's data

A company needs to train a foundation model (FM) on its business data. The dataset size is 1 GB. Training and evaluation must complete within three weeks. Which solution meets these requirements?

A. Retrieval-augmented generation (RAG)

C. Instruction-based fine-tuning

An insurance company is building an application that uses a large language model (LLM) for AI-powered document classification. It has a dataset of 10,000 documents with corresponding labels (e.g., 'insurance claim', 'invoice'). The model must learn the classification task and integrate knowledge from the entire dataset. Which solution minimizes prompt size while meeting these requirements?

D. Instruction-based fine-tuning

AWS AIF-C01 Chapter 3 Free Practice Test 2025 — AI Practitioner Exam Questions

AWS AIF-C01 Chapter 3 Practice Questions (100 Questions)

Practice questions for the AWS AIF-C01 (AI Practitioner) exam, Chapter 3.

Q1. A company wants to create a reusable prompt structure adaptable across different tasks and domains. Which prompt engineering technique satisfies these requirements?
- A. Zero-Shot Prompting
- B. Few-Shot Prompting
- C. Chain-of-Thought Prompting
- D. Prompt Templates
View question →
Q2. A company provides its large language model (LLM) with the following prompt: 'The cinematography of this movie is really terrible. // Negative sentiment. This movie has many exciting action scenes. // Positive sentiment. This movie is okay. // Neutral sentiment. How awful the cinematography is! //'. Which prompt technique does this describe?
- A. Few-Shot
- B. Zero-Shot
- C. Tree-of-Thought
- D. Chain-of-Thought
View question →
Q3. A law firm wants to use an Amazon Bedrock knowledge base as a retrieval-augmented generation (RAG) solution. The knowledge base contains years of regulatory requirements. The law firm needs a solution to research regulatory requirements that were valid during a given time period. Many documents ingested into the knowledge base do not contain any dates in their content. Which solution will minimize operational overhead to meet these requirements?
- A. Build and deploy an AWS Lambda function to request specific document content from the knowledge base
- B. Build and deploy containers on Amazon Elastic Container Service (ECS) to request specific document content from the knowledge base
- C. Build and deploy containers on Amazon Elastic Kubernetes Service (EKS) to request specific document content from the knowledge base
- D. Ingest metadata into the knowledge base and retrieve content only from documents tagged with specific dates
View question →
Q4. A data scientist wants to use a diffusion model to generate images from text prompts. The generated images are often blurry and contain unwanted noise. Which prompting technique can help remove unwanted noise?
- A. Chain-of-thought prompting
- B. Negative prompting
- C. Few-shot prompting
- D. Single-shot prompting with context
View question →
Q5. A company wants to build an internal chatbot application using Amazon Bedrock to answer employee questions based on internal documentation. The company aims to minimize application development costs. Which solution will minimize operational overhead to meet these requirements?
- A. Use fine-tuning to adapt an LLM with the company's data
- B. Use continual pre-training to adapt an LLM with the company's data
- C. Create a knowledge base and use retrieval-augmented generation (RAG)
- D. Train an LLM from scratch using only the company's data
View question →
Q6. A company seeks a solution that supports domain adaptation for ML models and provides a repository of pre-trained models and fine-tuning algorithms. Which Amazon SageMaker feature satisfies these requirements?
- A. SageMaker Autopilot
- B. SageMaker JumpStart
- C. SageMaker Model Registry
- D. SageMaker Feature Store
View question →
Q7. A company needs to train a foundation model (FM) on its business data. The dataset size is 1 GB. Training and evaluation must complete within three weeks. Which solution meets these requirements?
- A. Retrieval-augmented generation (RAG)
- B. Domain-adaptive fine-tuning
- C. Instruction-based fine-tuning
- D. Prompt engineering
View question →
Q8. An insurance company is building an application that uses a large language model (LLM) for AI-powered document classification. It has a dataset of 10,000 documents with corresponding labels (e.g., 'insurance claim', 'invoice'). The model must learn the classification task and integrate knowledge from the entire dataset. Which solution minimizes prompt size while meeting these requirements?
- A. Single-shot prompting
- B. Few-shot prompting
- C. Continual pre-training
- D. Instruction-based fine-tuning
View question →
Q9. A data scientist is using Amazon SageMaker to conduct text generation experiments with a large language model (LLM). The data scientist wants to evaluate whether the model exhibits bias related to gender, age, or race in its responses. Which type of evaluation satisfies these requirements?
- A. Factual knowledge
- B. Prompt stereotyping
- C. Toxicity
- D. Semantic robustness
View question →
Q10. A data scientist needs to monitor real-time unfair predictions from a model based on sensitive attributes such as race, gender, and age. The model is deployed on Amazon SageMaker. Which AWS service or feature satisfies this requirement?
- A. SageMaker Model Monitor
- B. SageMaker Clarify
- C. SageMaker Ground Truth
- D. SageMaker Data Wrangler
View question →
Q11. Which AWS service or feature can send low-confidence ML predictions to human reviewers for validation?
- A. Amazon CloudWatch Alarms
- B. AWS Lambda
- C. Amazon Augmented AI (A2I)
- D. Amazon SageMaker Experiments
View question →
Q12. What legal risk is associated with generative AI systems producing biased or discriminatory outputs?
- A. Antitrust violations
- B. Violations of consumer protection laws
- C. Data privacy violations
- D. Intellectual property (IP) infringement
View question →
Q13. A company is using a text-generation foundation model (FM) on Amazon Bedrock to power a customer-facing chatbot service. The company wants to prevent the FM from answering potentially sensitive or inappropriate customer questions. Which solution incurs the least operational overhead while meeting these requirements?
- A. Train a text classifier on Amazon SageMaker to detect and filter unwanted questions
- B. Define prohibited topics in Amazon Bedrock Guardrails and enable guardrails when invoking the FM
- C. Use Amazon Comprehend’s entity extraction capability to prompt and filter unwanted questions
- D. Train a sensitive-topic detection model using Amazon SageMaker BlazingText built-in algorithm
View question →
Q14. A company operating in a highly regulated industry wants to deploy a large language model (LLM) for content generation use cases. What is the primary challenge associated with human-centered LLM design in such environments?
- A. Low computational capacity of AI systems
- B. Model transparency and explainability
- C. Lack of predictive accuracy
- D. Large number of model parameters
View question →
Q15. A developer is integrating a generative AI-powered chatbot into an application and notices offensive language appearing in the application’s chat history. Which AWS service can the developer use to audit content and prevent offensive language?
- A. Amazon Macie
- B. Amazon Polly
- C. Amazon SageMaker Clarify
- D. Amazon Comprehend
View question →
Q16. A company is building a neural network-based model to classify text documents into multiple categories. The company wants to ensure stakeholders can interpret the model’s behavior and predictions. Which method incurs the least operational overhead while satisfying this requirement?
- A. Rely solely on model accuracy and performance metrics
- B. Manually inspect model weights and features to understand internal mechanisms
- C. Develop custom interpretability methods tailored to the selected model architecture
- D. Use model-agnostic interpretability methods, such as Shapley Additive Explanations (SHAP)
View question →
Q17. A company is implementing a generative AI solution and must ensure security protocols and governance frameworks meet regulatory requirements. These requirements help assess risks, governance, and controls for generative AI applications. Which service or framework satisfies these requirements?
- A. AWS Serverless Application Model (AWS SAM)
- B. Generative AI Security Scope Matrix
- C. AWS Generative AI Resilience Lifecycle Framework
- D. AWS Artifact
View question →
Q18. A company stores customer data in Amazon S3. Its compliance policy prohibits using any personally identifiable information (PII) or protected health information (PHI) for ML model training. Which AWS service can the company use to discover and monitor sensitive data?
- A. Amazon Inspector
- B. AWS Config
- C. AWS Audit Manager
- D. Amazon Macie
View question →
Q19. Which control improves security for AI and ML workloads?
- A. Use data encryption and the principle of least privilege to control access
- B. Establish an open computing and networking environment
- C. Use security questions to protect credentials
- D. Delete previous workloads to maintain a stable version
View question →
Q20. A company wants to train models using confidential data stored in Amazon S3. The company must prevent training data from reaching the internet. Which solution meets these requirements?
- A. Amazon Macie
- B. AWS PrivateLink
- C. AWS Identity and Access Management (IAM)
- D. Amazon Inspector
View question →
Q21. A company developing AI applications wants to obtain guidance on global standards and frameworks. The company wishes to ensure its AI applications comply with global standards. Which solution meets these requirements?
- A. AWS CloudTrail
- B. AWS Trusted Advisor
- C. International Organization for Standardization (ISO)
- D. Amazon Inspector
View question →
Q22. A company has designed a public chatbot powered by a large language model (LLM). The company is concerned about prompt injection attacks and wants to implement defenses against them. Which strategy most effectively prevents various types of prompt injection attacks?
- A. Write a system prompt describing topics users may ask about
- B. Use content filters to block inappropriate behavior and avoid criminal activity
- C. Use word filters to block keywords commonly associated with prompt injection attacks
- D. Use salted sequence tags to wrap instructions, such as (tag123)
View question →
Q23. A company’s governance team wants to track the risk rating of each large language model (LLM) used, without performing any product customization. The company uses Amazon SageMaker to host and deploy LLMs. Which SageMaker feature meets these requirements?
- A. SageMaker Model Registry
- B. SageMaker Model Cards
- C. SageMaker Clarify
- D. SageMaker Experiments
View question →
Q24. A company wants to apply single-layer object-level server-side encryption to data stored in an Amazon S3 bucket. Data scientists will use this data to train ML models. Data scientists want to control encryption key rotation. Which solution satisfies these requirements with minimal operational overhead?
- A. Client-side encryption using the Amazon S3 Encryption Client
- B. Server-side encryption with AWS KMS keys (SSE-KMS)
- C. Server-side encryption with customer-provided keys (SSE-C)
- D. Double-layer server-side encryption with AWS KMS keys (DSSE-KMS)
View question →
Q25. A financial services company wants to deploy an agent powered by a large language model (LLM) that provides investment account advice to customers and automatically executes transactions. From the list below, select the correct prompt engineering risk corresponding to the following scenario: An attacker forces the LLM to output all instructions it receives from the frontend application. This action reveals implementation vulnerabilities exploitable by the attacker.
- A. Extract conversation history
- B. Prompt persona switching
- C. Extract prompt template
- D. Ignore prompt template
View question →
Q26. A company wants to track API invocation metrics for its Amazon Bedrock models. Which AWS service meets this requirement?
- A. Amazon CloudWatch
- B. AWS Config
- C. AWS Artifact
- D. AWS Inspector
View question →
Q27. Select the correct AWS service or feature from the list below for the following task. One or more AWS services or features should be selected. Implement identity authentication and resource-level access control.
- A. Amazon Bedrock guardrails
- B. AWS Identity and Access Management (IAM)
View question →
Q28. A company wants to use generative AI to create product descriptions for its website. What limitations of generative AI should the company consider?
- A. Generative AI models may produce biased or inappropriate content, requiring human review and editing
- B. Generative AI cannot handle the large volume of data required for product descriptions
- C. Generative AI cannot generate multilingual text required for e-commerce websites
- D. Generative AI models lack the ability to understand and incorporate product specifications and details
View question →
Q29. A data scientist observes that a model achieves high accuracy on training data but low accuracy on test data. What phenomenon explains these results?
- A. Insufficient training time
- B. Underfitting
- C. Excessive training data
- D. Overfitting
View question →
Q30. A company wants to evaluate the performance of a foundation model (FM) used for text generation. Which technique or metric satisfies these requirements?
- A. Reinforcement learning
- B. F1 score
- C. Recall-Oriented Understudy for Gisting Evaluation (ROUGE)
- D. Fine-tuning
View question →
Q31. A travel company wants to use a pre-trained generative AI model to generate background images for marketing materials. The company lacks ML expertise and does not want to build or host ML models. Which AWS service satisfies these requirements?
- A. Amazon Bedrock
- B. Amazon SageMaker JumpStart
- C. Amazon Rekognition
- D. Amazon Personalize
View question →
Q32. In the context of generative AI, what is a foundation model (FM)?
- A. A task-specific model trained in a specialized domain (e.g., finance or medicine), serving as the foundational model for that domain
- B. A large, general-purpose model pre-trained on diverse datasets and adaptable to downstream tasks via fine-tuning
- C. A theoretical framework for understanding how different models learn representations
- D. A basic architecture serving as the starting point for designing more complex neural networks
View question →
Q33. An e-commerce company wants to generate personalized product descriptions for its online store. Which prompt engineering technique will satisfy these requirements with minimal operational effort?
- A. Few-shot prompting using well-written product description examples
- B. Zero-shot prompting without any examples
- C. Fine-tuning optimized based on customer engagement metrics
- D. Continued pre-training across different domains
View question →
Q34. A company wants to log API calls made to Amazon Bedrock. For compliance purposes, the logs must include API call details, the identity of the calling user, and the timestamp. Which AWS service satisfies these requirements?
- A. Amazon Inspector
- B. Amazon CloudWatch
- C. AWS Trusted Advisor
- D. AWS CloudTrail
View question →
Q35. A company wants to derive insights from diverse data sources and formats to improve its business operations. Data sources include call center audio, customer text feedback, product images, and scanned documents. Which AWS service can detect text and handwritten content from invoices stored in PNG format?
- A. Amazon Polly
- B. Amazon Textract
- C. Amazon Kendra
- D. Amazon Comprehend
View question →
Q36. A company wants to derive insights from diverse data sources and formats to improve its business operations. Data sources include call center audio, customer text feedback, product images, and scanned documents. Which solution will improve transcription accuracy for domain-specific language?
- A. Using a custom bot in Amazon Lex
- B. Using a custom language model in Amazon Translate
- C. Using batch language identification in Amazon Transcribe
- D. Using a custom language model in Amazon Transcribe
View question →
Q37. What is an effective data format for instruction-based fine-tuning?
- A. Images with category labels
- B. Curated playlists with recommended music
- C. Prompt-response text pairs
- D. Audio files with transcriptions
View question →
Q38. A company uses Amazon SageMaker for its ML models. The company wants to implement a solution enabling model owners to create model cards. Model cards should include intended use, risk rating, training details, and evaluation results. Which SageMaker feature satisfies these requirements?
- A. SageMaker Role Manager
- B. SageMaker Model Cards
- C. SageMaker Model Dashboard
- D. SageMaker Model Monitor
View question →
Q39. A company has a containerized frontend application for its AI application. The company must implement a solution to assess the security posture of its AWS environment. The solution must identify potential security vulnerabilities in the application’s Amazon EC2 instances and Amazon Elastic Container Registry (Amazon ECR) repositories and provide remediation recommendations. Which AWS service satisfies these requirements?
- A. AWS CloudTrail
- B. AWS Config
- C. AWS Inspector
- D. AWS Artifact
View question →
Q40. A company is deploying a solution on AWS to enhance the semantic search capability of its knowledge base. The company plans to integrate this solution with Amazon Bedrock. Which AWS service can be used to protect access to Amazon Bedrock?
- A. Amazon Macie
- B. Amazon Rekognition
- C. AWS Identity and Access Management (IAM)
- D. AWS Config
View question →
Q41. A financial company wants to use open-source foundation models (FMs) to evaluate whether loan agreements comply with regulatory rules. The company wants to reduce manual effort required during audits. Which AWS service satisfies these requirements?
- A. Amazon SageMaker JumpStart
- B. Amazon Textract
- C. Amazon Kendra
- D. Amazon Q Business
View question →
Q42. A company wants to improve the consistency and quality of responses from its large language model (LLM) by granting the model access to external knowledge sources. Which technique satisfies this requirement with minimal development effort?
- A. Fine-tuning
- B. Retrieval-Augmented Generation (RAG)
- C. In-context learning
- D. Prompt engineering
View question →
Q43. An e-commerce company plans to deploy an AI recommendation system for customer data. The data is stored in Amazon S3. The company wants to comply with data governance and privacy regulations and identify and protect personally identifiable information (PII). Which AWS service meets these requirements?
- A. AWS Config
- B. AWS CloudTrail
- C. AWS Trusted Advisor
- D. Amazon Macie
View question →
Q44. Which metric can evaluate foundational model (FM) performance in the context of text summarization?
- A. Recall
- B. F1 Score
- C. Classification Accuracy
- D. ROUGE-N
View question →
Q45. A company hosts its ML models on AWS. It requires documentation related to AWS security and compliance. Which AWS service meets this requirement?
- A. Amazon CloudWatch
- B. AWS CloudTrail
- C. AWS Artifact
- D. AWS Trusted Advisor
View question →
Q46. Which concept describes the development of algorithms and statistical models that enable computer systems to perform complex tasks without explicit instructions?
- A. Object-Oriented Programming (OOP)
- B. Amazon SageMaker Built-in Algorithms
- C. Inference
- D. ML
View question →
Q47. A company is exploring generative AI foundational models (FMs) to build a domain-specific text-based solution. The solution will be used internally to generate blog posts and marketing content, which will be published on the company’s external website. What is a disadvantage of using generative AI FMs to generate blog posts?
- A. Inaccuracy of generated content
- B. Response latency of generative AI FMs
- C. Ability to summarize information
- D. Reduced time-to-market
View question →
Q48. A company is exploring generative AI foundational models (FMs) to build a domain-specific text-based solution. The solution will be used internally to generate blog posts and marketing content, which will be published on the company’s external website. Which AWS service can the company use to select a generative AI FM?
- A. Amazon Augmented AI (A2I)
- B. Amazon Bedrock
- C. Amazon Lex
- D. Amazon Personalize
View question →
Q49. A company is exploring generative AI foundational models (FMs) to build a domain-specific text-based solution. The solution will be used internally to generate blog posts and marketing content, which will be published on the company’s external website. Which technique can the company use to ensure generated content includes domain-specific terminology?
- A. Domain-adaptive fine-tuning
- B. Reinforcement Learning from Human Feedback (RLHF)
- C. Instruction-based fine-tuning
- D. Supervised learning
View question →
Q50. A company is exploring generative AI foundational models (FMs) to build a domain-specific text-based solution. The solution will be used internally to generate blog posts and marketing content, which will be published on the company’s external website. Which method should the company use to assess the quality of generated content before publishing it online?
- A. Implementing guardrails
- B. Using Retrieval-Augmented Generation (RAG)
- C. Performing few-shot prompt engineering
- D. Incorporating human evaluation
View question →
Q51. A company is exploring the use of generative AI foundation models (FMs) to build a domain-specific, text-based solution. This solution will be used internally to generate blog posts and marketing content, which will be published on the company's external website. When using generative AI, what risk must the company consider?
- A. Increased operational costs
- B. Violation of data privacy regulations
- C. Intellectual property (IP) infringement claims
- D. Prompt injection attacks
View question →
Q52. A company is exploring the use of generative AI foundation models (FMs) to build a domain-specific, text-based solution. This solution will be used internally to generate blog posts and marketing content, which will be published on the company's external website. Which AWS service or framework can the company use to assess the level of ownership required for the AI solution?
- A. Generative AI Security Scoping Matrix
- B. AWS Trusted Advisor
- C. Amazon Inspector
- D. Amazon Macie
View question →
Q53. A global telecommunications company is building an AI-assisted tool to enhance customer chat capabilities. All generated chat responses must be positive, friendly, and unbiased. Which evaluation method will identify a foundation model (FM) that satisfies these requirements?
- A. Use Amazon SageMaker Clarify to identify bias in a historical customer chat dataset
- B. Review each model's sample benchmark performance using AWS AI Services Cards
- C. Use Amazon Augmented AI (A2I) to review responses before sending them to customers
- D. Use Amazon SageMaker Clarify to quantify model toxicity on a test dataset
View question →
Q54. A company is building a generative AI application for repetitive document summarization tasks. Initially, summaries generated by the foundation model (FM) vary in format and quality. Prompt engineering is being applied to develop prompt templates to improve consistency in quality and format. Which solution will improve output quality?
- A. Include links to the document repository containing the documents to be summarized
- B. Describe the heading format of information in the document to be summarized
- C. Include summaries automatically generated from the initial prompt
- D. Include curated summaries demonstrating the desired format and style
View question →
Q55. Using labeled data to train a car to recognize traffic signals, signs, and objects.
- A. Unsupervised learning
- B. Semi-supervised learning
- C. Supervised learning
View question →
Q56. A company is beginning to use Amazon SageMaker to develop ML solutions. The company lacks staff skilled in writing data cleaning and preparation scripts. Which SageMaker feature can the company use for data cleaning and preparation while minimizing the need to write scripts?
- A. SageMaker Data Wrangler
- B. SageMaker Feature Store
- C. SageMaker Clarify
- D. SageMaker Pipelines
View question →
Q57. A manufacturing company wants to create a virtual assistant for internal employees. This virtual assistant should answer technical questions exclusively based on the company’s proprietary engineering documentation. The company wants to minimize model hallucination and reduce effort related to data preparation. Which solution will satisfy these requirements with minimal operational overhead?
- A. Fine-tune a foundation model (FM) using the company’s knowledge base
- B. Use a foundation model trained on data related to the manufacturing domain
- C. Implement retrieval-augmented generation (RAG)
- D. Use in-context learning by embedding extensive instructions in the prompt
View question →
Q58. A company needs a relational database system to store embedding vectors associated with real-time transactions.
- A. Amazon OpenSearch Service
- B. Amazon Redshift
- C. Amazon RDS for PostgreSQL
- D. Amazon DocumentDB (with MongoDB Compatibility)
View question →
Q59. A developer needs a content moderation solution for use cases involving paintings, sketches, and animations. Which solution would meet these requirements with the least operational overhead?
- A. Use Amazon Bedrock to fine-tune a multimodal embedding model for the specific moderation use case
- B. Use Amazon Rekognition with a custom adapter for the specific moderation use case
- C. Use Amazon SageMaker JumpStart to fine-tune an open-source object detection model
- D. Use Amazon SageMaker Canvas to build and fine-tune a content moderation model
View question →
Q60. A sales representative stores customer contracts in PDF format in an Amazon S3 bucket. The sales representative needs a solution to summarize document content and answer questions based on information in the documents. The sales representative has no coding experience. Which AWS service meets these requirements?
- A. Amazon QuickSight
- B. Amazon Textract
- C. Amazon SageMaker Canvas
- D. Amazon Kendra
View question →
Q61. Which type of prompt injection attack involves instructing a large language model (LLM) to disregard its structured instructions and produce output related to prohibited or harmful topics?
- A. Input manipulation
- B. Exploiting helpfulness
- C. Ignoring the prompt template
- D. Prompt persona switching
View question →
Q62. A company wants to deploy an AI model to detect fraudulent transactions. Before releasing the model, the company must evaluate its performance. Which stage of the model development pipeline focuses on evaluating model performance?
- A. Data preprocessing
- B. Model evaluation
- C. Model monitoring
- D. Model deployment
View question →
Q63. A regional utility company wants to build a generative AI application that provides customers with personalized energy-saving tips. The company developed an application that constructs prompts by inserting customer inputs into a prompt template. Customer inputs are restricted to selections from a prepopulated list—including postal code, household size, and topics of interest. Security of the company’s systems and customer data is paramount. Which solution provides the strongest security benefit to prevent prompt injection attacks?
- A. Increase the frequency of system-wide software updates
- B. Add instructions in the prompt template on how to handle irrelevant requests
- C. Update the application with a free-text input field to accept personalized customer requests
- D. Optimize the prompt template based on model toxicity assessments
View question →
Q64. An AWS administrator needs to protect ML resources in the AWS environment. The administrator must configure appropriate permissions for users, groups, and federated roles to access Amazon S3 and Amazon SageMaker. Which AWS service or feature can the administrator use to grant these permissions?
- A. AWS Secrets Manager
- B. AWS Key Management Service (AWS KMS)
- C. S3 bucket policies
- D. AWS Identity and Access Management (IAM)
View question →
Q65. A company has collected hourly temperature readings over the past 30 days. The company wants to train an ML model to predict temperatures for the next 24 hours. Which approach satisfies this requirement?
- A. Clustering algorithm
- B. Anomaly detection algorithm
- C. Forecasting algorithm
- D. Classification algorithm
View question →
Q66. A company is building an ML application that evaluates customer feedback and classifies it as positive or negative. Which ML model should the application use?
- A. Binary classification model
- B. Multiclass classification model
- C. Image classification
- D. Regression model
View question →
Q67. A company wants to build a chatbot that can engage in natural language conversations with customers. The company wants to implement a managed solution. Which AWS service meets these requirements?
- A. Amazon Polly
- B. Amazon Transcribe
- C. Amazon Lex
- D. Amazon SageMaker
View question →
Q68. A data scientist is building an AI model to predict sales volume from market datasets. After ingesting the data, the data scientist wants to determine which market parameters most influence sales volume. Then, the data scientist will include those parameters in the model. Which step of the ML lifecycle does this describe?
- A. Data collection
- B. Feature selection
- C. Model validation
- D. Monitoring
View question →
Q69. An ML engineer has deployed a foundation model (FM) and wants to adapt the FM’s response format. The ML engineer wants to implement a technique to adapt the FM’s response format. Which technique will most cost-effectively meet these requirements?
- A. Retrieval-Augmented Generation (RAG)
- B. Prompt engineering
- C. Feature engineering
- D. Fine-tuning
View question →
Q70. A company wants to use Amazon Bedrock to create document drafts. What should be the company’s first step?
- A. Select an input token quota for generation
- B. Select a foundation model
- C. Select a vector database
- D. Select an embedding for generation
View question →
Q71. A data scientist is studying a text generation model that uses embeddings to represent content as vectors. What is the purpose of using embeddings in this context?
- A. Tokenize input text into individual words or subword units
- B. Encode input text into a sequence of numeric tokens
- C. Apply attention mechanisms to input text
- D. Capture semantic relationships between words based on word vector representations
View question →
Q72. A data scientist passes the phrase 'the movie was really great' to an embedding model. What is the model’s output?
- A. A word representing the phrase’s sentiment
- B. A value assessing the phrase’s credibility
- C. An image depicting the phrase
- D. A numeric array
View question →
Q73. Developing a large-scale English-to-French patent repository that includes image processing.
- A. Traditional ML models
- B. Generative AI models
View question →
Q74. A developer is building a web application for podcast hosts to summarize podcast episodes. First, podcast hosts must upload transcripts of podcast episodes to the web application. Then, podcast hosts receive an email containing two summaries of the podcast episode. Podcast hosts will use these summaries to promote content on social media. Which AWS service meets these requirements?
- A. Amazon Textract
- B. Amazon Transcribe
- C. Amazon Comprehend
- D. Amazon Bedrock
View question →
Q75. A company wants to create a virtual assistant to support customer queries. The virtual assistant will provide answers in natural language. These answers will be based on documents stored in Amazon S3. The company wants to use a managed service and does not want to configure parameters for models, embeddings, conversational capabilities, or vector storage. Which solution best meets these requirements in the most operationally efficient way?
- A. Amazon Comprehend
- B. Amazon Q Business
- C. Amazon SageMaker JumpStart
- D. Amazon Lex
View question →
Q76. A news company wants to increase click-through rates by delivering more relevant news to each subscriber. Which AWS service meets these requirements with the least operational overhead?
- A. Amazon Comprehend
- B. Amazon Personalize
- C. Amazon Kendra
- D. Amazon Bedrock
View question →
Q77. A company wants to develop an internal solution to summarize multiple emails and lengthy meeting notes. The company plans to use foundation models (FMs) on Amazon Bedrock to implement this solution. When selecting a model, which model attribute should be prioritized first?
- A. Temperature
- B. Top P
- C. Stop sequence
- D. Context window
View question →
Q78. A university wants to build an application using foundation models (FMs) to help students solve math problems from practice exams. The application should provide reasoning and step-by-step explanations to help students understand why answers are correct or incorrect. Which technique most reliably enhances FM capabilities for this use case?
- A. Single-shot prompting
- B. Few-shot prompting
- C. Retrieval-Augmented Generation (RAG)
- D. Chain-of-Thought (CoT)
View question →
Q79. What is the primary purpose of using prompt templates in prompt engineering?
- A. To guide the model by breaking down complex problems into smaller logical parts
- B. To train the model to generalize and make predictions on unseen categories or tasks
- C. To provide consistent and structured input and output formats to the model
- D. To teach the model new categories or tasks using limited data
View question →
Q80. A medical device company has added AI-generated product summaries to its online product catalog. The company wants to incorporate industry-specific terminology to improve output quality. An ML team has access to large volumes of unlabeled industry-specific standards and research documents. Which ML technique satisfies these requirements?
- A. Using prompt engineering to include edited examples in the next prompt
- B. Using Retrieval-Augmented Generation (RAG) to include relevant terminology
- C. Customizing the foundation model via continued pretraining on the dataset
- D. Customizing the foundation model via instruction fine-tuning on an industry dataset
View question →
Q81. A company wants to use a large language model (LLM) to learn the language specific to its industry. The company has large volumes of unlabeled data containing industry-specific language. Which solution meets these requirements with the least operational overhead?
- A. Fine-tuning the LLM using the company’s data
- B. Continuing pretraining of the LLM using the company’s data
- C. Training a new LLM from scratch using the company’s data
- D. Providing the company’s data as context in the LLM’s prompts
View question →
Q82. A company is developing an ML model to generate natural-language responses for a customer service chatbot. It needs to evaluate how similar the model’s generated responses are to subject-matter expert (SME) responses. The company has a dataset of SME-validated question-answer pairs. Which metric should the company use to evaluate model performance?
- A. BERTScore
- B. Mean Squared Error (MSE)
- C. Perplexity
- D. F1 Score
View question →
Q83. A financial company is developing a machine learning (ML) model to predict customer loan eligibility. The company wants a solution that can reliably identify bias in training data and model predictions. Which AWS service meets these requirements?
- A. Amazon Rekognition
- B. Amazon SageMaker Clarify
- C. Amazon SageMaker Ground Truth
- D. Amazon SageMaker Model Monitor
View question →
Q84. An AI specialist is training a regression ML model and observes bias and variance during training. Which bias-variance pattern indicates model overfitting?
- A. Low bias, low variance
- B. Low bias, high variance
- C. High bias, low variance
- D. High bias, high variance
View question →
Q85. When a model performs well on training data but fails to generalize to unseen data.
- A. Bias
- B. Fairness
- C. Overfitting
- D. Underfitting
- E. Interpretability
View question →
Q86. A company applies human-centered design to its AI application using Reinforcement Learning from Human Feedback (RLHF). The company wants to create a trustworthy training dataset that incorporates human feedback to improve its large language model (LLM) under development. Which solution meets these requirements?
- A. Use Amazon SageMaker built-in algorithms
- B. Use Amazon SageMaker Ground Truth
- C. Use Amazon SageMaker Autopilot
- D. Use Amazon SageMaker Pipelines
View question →
Q87. A company needs to provide information about its custom ML models for audit purposes. This information must include model training and performance details. Which AWS service or feature meets this requirement?
- A. Amazon Bedrock Agents
- B. Amazon Bedrock Guardrails
- C. Amazon SageMaker Model Cards
- D. Amazon SageMaker Ground Truth
View question →
Q88. A healthcare company must comply with regulatory policies mandating that data stored in Amazon S3 must not traverse the public internet when transferred to Amazon EC2 instances for ML model training. Which solution meets these requirements?
- A. Gateway endpoint
- B. S3 Access Point
- C. VPC Flow Logs
- D. Transit Gateway
View question →
Q89. A social media company anticipates a significant increase in creative content types on its platform—including photos, animations, and illustrations. The company wants to perform content moderation on these creative content types. Which solution satisfies these requirements with minimal operational overhead?
- A. Fine-tune the Amazon Titan model in Amazon Bedrock
- B. Build a content moderation model using Amazon SageMaker Canvas
- C. Build a content moderation model using Amazon SageMaker Studio
- D. Use content moderation in Amazon Rekognition
View question →
Q90. A company hosts several web applications on Amazon EC2 instances. These applications need to access foundation models (FMs) in Amazon Bedrock. For audit purposes, the company wants to automatically log all web application access to Amazon Bedrock. The mechanism must capture user, role, and timestamp for every API invocation and inference call. Which AWS service meets these requirements?
- A. Amazon CloudWatch
- B. Amazon Macie
- C. AWS Trusted Advisor
- D. AWS CloudTrail
View question →
Q91. A company's users develop ML models in Amazon SageMaker Canvas. After development, these models need to be reviewed and approved by the data science team working in SageMaker Studio. Which AWS service enables the least operational overhead to provide data scientists with access to the ML models?
- A. SageMaker JumpStart
- B. SageMaker Model Registry
- C. Amazon DynamoDB
- D. Amazon Elastic Container Registry (ECR)
View question →
Q92. A lending company is building an AI-powered generative solution to offer discounts to loan applicants based on specific business criteria. The company wants to build and use AI models responsibly, minimizing potential bias that could negatively impact certain customers. Which actions should the company take to meet these requirements? (Choose TWO.)
- A. Detect imbalances or disparities in the data.
- B. Ensure the model runs frequently.
- C. Evaluate model behavior to provide transparency to stakeholders.
- D. Use recall-oriented Gisting evaluation of foundational research (ROUGE) techniques to ensure 100% model accuracy.
- E. Ensure model inference time remains within acceptable limits.
View question →
Q93. A healthcare company is developing a foundation model (FM) for diagnostics. The company requires the model to be transparent and interpretable to meet regulatory requirements. Which solution satisfies these requirements?
- A. Use Amazon Inspector to configure security and compliance.
- B. Use Amazon SageMaker Clarify to generate simple metrics, reports, and examples.
- C. Use Amazon Macie to encrypt and protect training data.
- D. Collect more data. Use Amazon Rekognition to add custom labels to the data.
View question →
Q94. A digital device company wants to forecast customer demand for memory hardware. The company has no coding expertise or knowledge of ML methodologies and therefore needs to develop a data-driven forecasting model. The company needs to analyze both internal and external data. Which solution meets these requirements?
- A. Store data in Amazon S3. Use Amazon SageMaker built-in algorithms with data from Amazon S3 to create an ML model and demand forecast.
- B. Import data into Amazon SageMaker Data Wrangler. Use SageMaker built-in algorithms to create an ML model and demand forecast.
- C. Import data into Amazon SageMaker Data Wrangler. Use the Amazon Personalize Trending-Now recipe to create an ML model and demand forecast.
- D. Import data into Amazon SageMaker Canvas. Build an ML model and demand forecast by selecting values from the data within SageMaker Canvas.
View question →
Q95. An AI developer is using an Amazon Bedrock foundation model to summarize customer service team meeting chats. The AI developer wants to store invocation logs to monitor model input and output data. Which strategy should the AI developer use?
- A. Configure AWS CloudTrail as the logging destination for the model.
- B. Enable invocation logging in Amazon Bedrock.
- C. Configure AWS Audit Manager as the logging destination for the model.
- D. Configure model invocation logging in Amazon EventBridge.
View question →
Q96. A company needs to build its own large language model (LLM) exclusively using its private data. The company is concerned about the environmental impact of training. Which Amazon EC2 instance type minimizes environmental impact during LLM training?
- A. Amazon EC2 C series
- B. Amazon EC2 G series
- C. Amazon EC2 P series
- D. Amazon EC2 Trn series
View question →
Q97. An accounting firm wants to implement a large language model (LLM) to automate document processing. The firm must do so responsibly to avoid potential risks. What should the company do when developing and deploying the LLM? (Choose TWO.)
- A. Include fairness metrics in model evaluation.
- B. Adjust the model’s temperature parameter.
- C. Modify training data to reduce bias.
- D. Avoid overfitting the training data.
- E. Apply prompt engineering techniques.
View question →
Q98. A company is building a contact center application and wants to derive insights from customer conversations. The company wants to analyze and extract key information from customer call audio. Which solution meets these requirements?
- A. Use Amazon Lex to build a conversational chatbot.
- B. Use Amazon Transcribe to transcribe call recordings.
- C. Use Amazon SageMaker Model Monitor to extract information from call recordings.
- D. Use Amazon Comprehend to create classification labels.
View question →
Q99. A company wants to deploy a conversational chatbot to answer customer questions. The chatbot is built using an Amazon SageMaker JumpStart model. The application must comply with multiple regulatory frameworks. Which features can the company demonstrate compliance for? (Choose two.)
- A. Auto-scaling inference endpoints
- B. Threat detection
- C. Data protection
- D. Cost optimization
- E. Loosely coupled microservices
View question →
Q100. An e-commerce company wants to build a solution to determine customer sentiment based on written product reviews. Which AWS services meet these requirements? (Choose two.)
- A. Amazon Lex
- B. Amazon Comprehend
- C. Amazon Polly
- D. Amazon Bedrock
- E. Amazon Rekognition
View question →