A company stores customer data in OpenSearch. The company wants an AI solution to retrieve specific customer information from the stored data. The AI solution must convert queries into data requests and generate CSV files from the results. Then, the AI solution must upload the CSV files to Amazon S3.
Which solution will meet these requirements in the MOST operationally-efficient way?
A. Create an AI agent to perform the required steps.
B. Use a single foundation model (FM) with few-shot prompting.
C. Create a software application without using AI to perform the required steps.
D. Train a decision tree model to generate a solution based on user questions.
Show Answer
Correct Answer: A
Explanation: An AI agent is designed to orchestrate multi-step workflows: interpret user requests, translate them into data retrieval operations against OpenSearch, invoke tools or APIs, generate CSV output from retrieved data, and upload the resulting files to Amazon S3. This is more operationally efficient than relying on prompting alone, building a custom application without AI orchestration, or using a decision tree model, which is not suitable for natural language query understanding and workflow automation.
Question 105
A company that streams media is selecting an Amazon Nova foundation model (FM) to process documents and images. The company is comparing Nova Micro and Nova Lite. The company wants to minimize costs.
Which model characteristics should the company consider to meet these requirements?
A. Nova Micro uses transformer-based architectures. Nova Lite does not use transformer-based architectures.
B. Nova Micro supports only text data. Nova Lite is optimized for numerical data.
C. Nova Micro supports only text. Nova Lite supports images, videos, and text.
D. Nova Micro runs only on CPUs. Nova Lite runs only on GPUs.
Show Answer
Correct Answer: C
Explanation: Amazon Nova Micro is a text-only model designed for low-cost text workloads. Amazon Nova Lite is a multimodal model that accepts text, images, and video, making the key distinction relevant to processing documents and images while considering cost. The other options incorrectly describe the architectures or hardware requirements.
Question 106
A research company is growing microbiological cultures. The company captures images of the cultures without any prior labeled data about growth areas. The company needs to identify the regions of the images that show culture growth.
Which ML technique will meet these requirements?
A. Logistic regression
B. Decision tree
C. Clustering
D. Dimensionality reduction
Show Answer
Correct Answer: C
Explanation: The data is unlabeled and the goal is to identify regions with similar characteristics corresponding to culture growth. This is an unsupervised learning problem, and clustering groups similar pixels or image regions without requiring labeled training data. Logistic regression and decision trees are supervised methods requiring labels, while dimensionality reduction primarily compresses or visualizes data rather than segmenting regions.
Question 107
A company uses Amazon SageMaker AI to generate article summaries in multiple languages. The company needs a metric to evaluate the quality of the summary translations in multiple languages.
Which evaluation metric will meet these requirements?
A. Recall-Oriented Understudy for Gisting Evaluation (ROUGE)
B. Bilingual evaluation understudy (BLEU)
C. Area Under the ROC Curve (AUC)
D. Precision
Show Answer
Correct Answer: B
Explanation: BLEU is the standard automatic evaluation metric for machine translation quality, comparing generated translations against one or more reference translations across languages. ROUGE is primarily intended for evaluating summarization quality, while AUC and Precision are classification metrics.
Question 108
A company is developing a product recommendation application by using a generative AI model. The company must minimize the application’s environmental impact.
Which solution will meet these requirements?
A. Optimize the deployed model architecture to prioritize computational efficiency during model inference.
B. Adopt a distributed inference approach by using multiple smaller models across multiple Availability Zones.
C. Adopt a hybrid strategy by deploying the model on premises and storing the data on AWS.
D. Deploy multiple models and use a dynamic model selection mechanism that queries different models randomly.
Show Answer
Correct Answer: A
Explanation: Optimizing the deployed model architecture for computational efficiency reduces compute, energy consumption, and resource usage during inference, aligning with the AWS Well-Architected Sustainability Pillar. The other options either increase infrastructure overhead, do not inherently reduce energy use, or add unnecessary model deployments and random selection.
Question 109
A company is building a job recommendation system based on job posting data and job seeker user profiles. The system shows bias in job recommendations based on gender for user profiles that are otherwise equivalent.
Which principle should the company follow to address this issue, according to AWS best practices for responsible AI?
A. Governance
B. Explainability
C. Controllability
D. Fairness
Show Answer
Correct Answer: D
Explanation: The correct answer is D. Fairness. AWS responsible AI principles include fairness, which focuses on identifying and mitigating unwanted bias so that similar individuals are treated equitably regardless of protected characteristics such as gender. Because equivalent user profiles receive different job recommendations based on gender, the issue is one of fairness rather than explainability, controllability, or governance.
Question 110
A company is using a large language model (LLM) to create a generative AI assistant. The company must choose an AI technique to ensure that the AI assistant generates the most factually correct responses. The company selects the Retrieval Augmented Generation (RAG) technique.
Which limitation of LLMs is the company trying to reduce?
A. Hallucinations
B. Security
C. Nondeterminism
D. Interpretability
Show Answer
Correct Answer: A
Explanation: Retrieval Augmented Generation (RAG) retrieves relevant information from external knowledge sources and provides it to the language model during generation. This grounds responses in factual data and primarily reduces hallucinations—confident but incorrect or fabricated answers. It does not primarily solve security, nondeterminism, or interpretability.
Question 111
A company has trained a foundation model (FM) to perform a specific task. The company needs to fine-tune the FM to perform a different but related task.
Which fine-tuning method will meet this requirement?
A. Hyperparameter tuning
B. Pre-training
C. Transfer learning
D. Reinforcement learning
Show Answer
Correct Answer: C
Explanation: Transfer learning starts with a pre-trained foundation model and fine-tunes it for a different but related task by reusing learned representations. Hyperparameter tuning only adjusts training settings, pre-training is the initial training phase rather than fine-tuning, and reinforcement learning is a different training paradigm used for learning from rewards.
Question 112
What is the primary purpose of system prompts in generative AI applications?
A. To authenticate user credentials to access responses from AI
B. To generate random responses by AI
C. To define the role and behavioral boundaries of AI
D. To compress output data from AI
Show Answer
Correct Answer: C
Explanation: System prompts provide the model with foundational instructions that define its role, behavior, tone, and constraints throughout the interaction. They are not used for authentication, generating random responses, or compressing output data.
Question 113
A company wants its AI models to be transparent and explainable.
Which combination of Amazon SageMaker AI features will meet these requirements? (Choose two.)
A. SageMaker Model Cards
B. SageMaker Pipelines
C. SageMaker Clarity
D. SageMaker Model Monitor
E. SageMaker Debugger
Show Answer
Correct Answer: A, C
Explanation: SageMaker Model Cards improve transparency by documenting model details, intended use, performance, risks, and governance information. SageMaker Clarify (the service is named Clarify, not Clarity) provides explainability through feature attribution (such as SHAP) and bias detection. Pipelines orchestrate workflows, Model Monitor detects data/model drift in production, and Debugger helps identify training issues rather than providing transparency or explainability.
$19
Get all 443 questions with detailed answers and explanations
Instant download HTML + PDF delivered the moment payment clears.
Secure Stripe checkout we never see or store your card details.
7-day refund if files are defective see our refund policy.