Fairness refers to the absence of bias in AI models. Using non-representative datasets leads to biased predictions, affecting specific demographics unfairly. Explainability, privacy, and transparency are important but not directly related to this scenario. References: AWS Responsible AI Framework.

Amazon SageMaker Feature Store is the correct solution for sharing and managing variables (features) across multiple teams during model development.

Amazon SageMaker Feature Store:

A fully managed repository for storing, sharing, and managing machine learning features across different teams and models.

It enables collaboration and reuse of features, ensuring consistent data usage and reducing redundancy.

Why Option A is Correct:

Centralized Feature Management: Provides a central repository for managing features, making it easier to share them across teams.

Collaboration and Reusability: Improves efficiency by allowing teams to reuse existing features instead of creating them from scratch.

Why Other Options are Incorrect:

B. SageMaker Data Wrangler: Helps with data preparation and analysis but does not provide a centralized feature store.

C. SageMaker Clarify: Used for bias detection and explainability, not for managing variables across teams.

D. SageMaker Model Cards: Provide model documentation, not feature management.

Creating an Amazon Bedrock knowledge base allows the integration of external or private data sources with a foundation model (FM) like Amazon Titan. This integration helps supplement the model with relevant data from the company ' s private data sources to enhance its responses.

Option C (Correct): " Create an Amazon Bedrock knowledge base " : This is the correct answer as it enables the company to incorporate private data into the FM to improve its effectiveness.

Option A: " Use a different FM " is incorrect because it does not address the need to supplement the current model with private data.

Option B: " Choose a lower temperature value " is incorrect as it affects output randomness, not the integration of private data.

Option D: " Enable model invocation logging " is incorrect because logging does not help in supplementing the model with additional data.

AWS AI Practitioner References:

Amazon Bedrock and Knowledge Integration: AWS explains how creating a knowledge base allows Amazon Bedrock to use external data sources to improve the FM’s relevance and accuracy.

Providing labeled data with both a prompt field and a completion field is the correct strategy for fine-tuning a foundation model (FM) on Amazon Bedrock.

Fine-Tuning Strategy:

To fine-tune a model, labeled data that pairs input prompts with the correct outputs (completions) is necessary.

This allows the model to learn the desired behavior or response style based on the provided examples.

Why Option A is Correct:

Proper Training Format: The prompt-completion pairs provide the necessary format for training the model to produce accurate outputs.

Customization: Ensures that the model is fine-tuned to the specific requirements of the company’s data and desired outputs.

Why Other Options are Incorrect:

B. Prepare a .txt file in .csv format: This does not align with the specific need for labeled data with prompts and completions.

C. Purchase Provisioned Throughput: Relates to read/write capacity in databases, not to model fine-tuning.

D. Train on journals and textbooks: Lacks the specific format and labeling required for fine-tuning.

This scenario demonstrates intellectual property (IP) infringement, where an AI-generated artifact (a cloned voice) uses the identifiable traits of a person without permission. According to AWS Responsible AI practices, using a person’s likeness, image, or voice in generative AI systems requires explicit rights and consent. Foundation models trained or applied without respecting these ownership boundaries violate ethical and legal frameworks under IP law. The AWS documentation emphasizes that generative AI must respect copyrighted or trademarked works and human likenesses. Responsible deployment requires consent-based data usage, license tracking, and content watermarking to prevent misappropriation. AWS Bedrock enforces such compliance principles through Guardrails and policy-based moderation to ensure content generation aligns with ethical and lawful standards.

Referenced AWS AI/ML Documents and Study Guides:

AWS Responsible AI Guidelines – Intellectual Property and Ethical Use

Amazon Bedrock Documentation – Responsible Use of Generative AI

Generative Pre-trained Transformers (GPT) are suitable for building an AI-based application that can generate SQL queries from natural language input provided by employees.

GPT for Natural Language Processing:

GPT models are designed for understanding and generating human-like text based on natural language input.

They can be fine-tuned to interpret specific tasks, such as converting natural language queries into SQL queries.

Why Option A is Correct:

Natural Language Understanding: GPT is highly effective for tasks that require understanding of human language and generating structured outputs like SQL.

User-Friendly: Requires minimal technology experience from employees, as they provide simple text input.

Why Other Options are Incorrect:

B. Residual neural network: Typically used in computer vision tasks, not for natural language-to-SQL conversion.

C. Support vector machine: Used for classification tasks, not for generating structured queries from text.

D. WaveNet: A deep generative model for audio data, unrelated to text-to-SQL tasks.

Few-shot learning involves providing a model with a few examples (shots) to learn from. For improving intent detection accuracy in a chatbot using a large language model (LLM), the data should consist of pairs of user messages and their corresponding correct intents.

Few-shot Learning for Intent Detection:

Few-shot learning aims to enable the model to learn from a small number of examples. For intent detection, the model needs to understand the relationship between user messages and the intended action or meaning.

Providing examples of user messages and the correct user intents allows the model to learn patterns in the phrasing or language that corresponds to each intent.

Why Option C is Correct:

User Messages and Intents: These examples directly teach the model how to map a user’s input to the appropriate intent, which is the goal of intent detection in chatbots.

Improves Accuracy: By using few-shot learning with these examples, the model can generalize better from limited data, improving intent detection.

Why Other Options are Incorrect:

A. Pairs of chatbot responses and correct user intents: Incorrect because it does not focus on user input but rather on outputs.

B. Pairs of user messages and correct chatbot responses: This would be useful for response generation, not intent detection.

D. Pairs of user intents and correct chatbot responses: Again, this is not aligned with detecting intents but with generating responses.

The verified answer is C. Use the LLM-as-a-judge approach with a rubric across prompts to compare model outputs. The question asks for a scalable way to compare two foundation models for chat summarization based on correctness and completeness. AWS documentation for Amazon Bedrock model evaluation states that an evaluation job using a judge model uses an LLM to score another model’s responses and provide an explanation of how it scored each prompt and response pair. This is specifically designed to compare model outputs across prompt datasets at scale, without requiring every output to be manually scored by humans.

AWS Bedrock documentation also lists built-in metrics for model evaluation jobs that use an LLM as a judge. The valid built-in metrics include Correctness and Completeness, which directly match the evaluation criteria in the question. Therefore, using LLM-as-a-judge with a rubric across prompts is the best fit because it evaluates the actual summaries produced by each model against the desired quality dimensions.

Option A is partially valid as an evaluation method, but it is not the most scalable approach. Human reviewers can provide high-quality judgments, but manual review becomes expensive and slow when comparing many prompts, outputs, and models. Option B is incorrect because latency and throughput measure operational performance, not whether a summary is correct or complete. A fast model can still produce incomplete or inaccurate summaries. Option D is incorrect because external benchmark rankings are generic. They do not prove that a model performs best on the company’s specific chat summarization workload.

The decisive terms are “scalable,” “correctness,” and “completeness.” Amazon Bedrock LLM-as-a-judge evaluation directly supports those dimensions, so option C is correct.

The correct answer is C because Amazon Bedrock ' s model evaluation feature allows users to compare outputs from different foundation models using human evaluation or automatic metrics. It enables the creation of structured evaluations where human reviewers (in this case, scientists) can assess model responses based on custom criteria like relevance, coherence, or accuracy.

From AWS documentation:

" Amazon Bedrock provides model evaluation capabilities that support both automatic and human evaluation. You can define custom evaluation prompts and collect assessments from reviewers to compare foundation model outputs for tasks such as summarization, text generation, and more. "

This solution is ideal for research workflows requiring domain experts to provide feedback on LLM-generated content.

Explanation of other options:

A. Amazon Personalize is used for building recommendation systems, not for evaluating model output.

B. Amazon Rekognition is used for analyzing images and videos (e.g., moderation, facial recognition), not textual output.

D. Amazon Comprehend provides NLP services like sentiment analysis, but sentiment is not sufficient for full quality evaluation of research paper generation.

Referenced AWS AI/ML Documents and Study Guides:

Amazon Bedrock Developer Guide – Model Evaluation Overview

AWS Generative AI Best Practices

AWS ML Specialty Study Guide – Evaluation and Feedback Loops in LLMs

The verified answer is A. Continuous pre-training. The company wants to expose an existing LLM to medical papers so the model becomes more familiar with medical topics, terminology, and writing patterns. AWS generative AI documentation defines continued pre-training as extending the pre-training phase of a foundation model by exposing it to additional unlabeled text from specific domains or corpora. It also explains that continued pre-training helps the model acquire deeper knowledge of new domains and adapt to domain-specific terminology and content types.

That exactly matches the scenario. Medical papers are typically domain text. The goal is not to train the model to follow a labeled input-output task, but to improve its domain familiarity and medical-language fluency. AWS security reference guidance also distinguishes continued pre-training from fine-tuning by stating that continued pre-training uses unlabeled data to enhance domain knowledge, while fine-tuning uses labeled data to optimize task-specific performance.

Supervised learning is incorrect because it requires labeled training examples, usually input-output pairs. The question only says the company has medical papers, not labeled prompts and responses.

Reinforcement learning is incorrect because it optimizes behavior based on rewards, feedback, or preference signals. The question does not describe reward modeling or human preference feedback.

Boosting is incorrect because boosting is an ensemble technique that combines weak learners to improve predictive performance. It is not the standard method for adapting an LLM to a specialized knowledge domain.

Because the company wants to familiarize an LLM with medical subject matter through medical papers, the correct technique is continuous pre-training.