A company stores customer personally identifiable information (PII) data. The company must store the PII data within the company's AWS Region.
Which aspect of governance does this describe?
A. Data mining
B. Data residency
C. Pre-training bias
D. Geolocation routing
Show Answer
Correct Answer: B
Explanation: The requirement that customer PII must be stored within a specific AWS Region concerns where data is physically or logically located to meet regulatory or policy constraints. This is the definition of data residency, not data mining, model bias, or routing behavior.
Question 6
HOTSPOT
-
A company is building an AI assistant application. The company must implement a core governance process for the application development project. The company must ensure that the application aligns with responsible AI practices.
Select and order the steps from the following list to correctly describe the implementation of a core governance process for this use case. Select each step one time.
Show Answer
Correct Answer: Step 1: Determine governance goals, risks, and policies.
Step 2: Put together a cross-functional AI governance group.
Step 3: Set up model monitoring mechanisms.
Explanation: Governance starts by defining objectives, risks, and policies. A cross‑functional group is then formed to own and enforce these rules. Ongoing model monitoring is implemented last to ensure compliance and responsible AI behavior during operation.
Question 7
A company wants to develop an interpretable ML model to assess the risk of loan applications.
Which type of ML model or algorithm will meet these requirements?
A. Deep learning model
B. Logistic regression model
C. K-means algorithm
D. Random cut forest algorithm
Show Answer
Correct Answer: B
Explanation: An interpretable model is required for loan risk assessment. Logistic regression provides transparent coefficients that show how each feature influences the prediction, produces clear probability outputs, and is widely accepted in regulated domains. The other options are either black-box (deep learning), unsupervised (K-means), or designed for anomaly detection (random cut forest).
Question 7
A company is using a large collection of web data to produce a large language model (LLM). The company completes a random initialization of the model’s weights. Next, the company fits the model to the data through a language objective modelling function.
Which stage of the model training process does this scenario describe?
A. Fine-tuning
B. Pre-training
C. Model selection
D. Deployment
Show Answer
Correct Answer: B
Explanation: The scenario describes initializing model weights randomly and training the model on a large corpus using a language modeling objective. This is the definition of pre-training, where a general-purpose LLM learns broad language patterns from large-scale data before any task-specific fine-tuning.
Question 8
HOTSPOT
-
A company wants to build generative AI applications by using Amazon Bedrock. The company wants to minimize development effort.
Select and order the model development techniques from the following list from the LEAST development effort to the MOST development effort. Each model development technique should be selected one time.
Explanation: Prompt engineering uses foundation models as-is with minimal setup.
RAG adds a retrieval layer and data indexing without retraining.
Fine-tuning requires curated data and training jobs.
Continued pre-training extends training on large corpora and is the most effort-intensive.
Question 8
A company uses an Amazon Bedrock large language model (LLM) in an application. During testing, the company observes different outputs from the same input.
What is the MOST likely cause of this issue?
A. The LLM is acting in a nondeterministic way.
B. The guardrails of the LLM are not configured properly.
C. The LLM has security vulnerabilities.
D. The LLM is acting in a deterministic way.
Show Answer
Correct Answer: A
Explanation: Large language models typically include stochastic elements (such as temperature and sampling), which can cause different outputs for the same input. Observing varying responses under identical inputs is most commonly due to the model’s nondeterministic behavior, not guardrails, security issues, or deterministic operation.
Question 9
A company is using a pre-trained large language model (LLM). The LLM must perform multiple tasks that require specific domain knowledge. The LLM does not have information about several technical topics in the domain. The company has unlabeled data that the company can use to fine-tune the model.
Which fine-tuning method will meet these requirements?
A. Full training
B. Supervised fine-tuning
C. Continued pre-training
D. Retrieval Augmented Generation (RAG)
Show Answer
Correct Answer: C
Explanation: The company has unlabeled domain-specific data and needs to add missing technical knowledge across multiple tasks. Continued pre-training adapts a pre-trained LLM to a specific domain using large amounts of unlabeled text, expanding its knowledge base. Supervised fine-tuning requires labeled data, full training is unnecessary and costly, and RAG retrieves external knowledge without embedding it into the model.
Question 9
A company wants to improve a large language model (LLM) for content moderation within 3 months. The company wants the model to moderate content according to the company's values and ethics. The LLM must also be able to handle emerging trends and new types of problematic content.
Which solution will meet these requirements?
A. Conduct continuous pre-training on a large amount of text-based internet content.
B. Create a high quality dataset of historical moderation decisions.
C. Fine-tune the LLM on a diverse set of general ethical guidelines from various sources.
D. Conduct reinforcement learning from human feedback (RLHF) by using real-time input from skilled moderators.
Show Answer
Correct Answer: D
Explanation: RLHF with skilled moderators directly aligns the model’s behavior with the company’s values and ethics, can be implemented within a short timeline, and continuously adapts to emerging trends and new forms of problematic content through real-time human feedback.
Question 10
A company wants to implement a generative AI solution to improve its marketing operations. The company wants to increase its revenue in the next 6 months.
Which approach will meet these requirements?
A. Immediately start training a custom FM by using the company's existing data.
B. Conduct stakeholder interviews to refine use cases and set measurable goals.
C. Implement a prebuilt AI assistant solution and measure its impact on customer satisfaction.
D. Analyze industry AI implementations and replicate the most successful features.
Show Answer
Correct Answer: B
Explanation: To increase revenue within a short 6‑month timeframe, the company should first ensure the generative AI initiative is tightly aligned with business goals. Conducting stakeholder interviews helps refine high‑impact marketing use cases, align expectations, and define measurable KPIs. This de‑risks implementation and enables faster, results‑driven deployment compared with training custom models or copying external solutions.
Question 10
Which outcome is a result of increasing model transparency?
A. Reduced need for model validation steps
B. Elimination of regulatory compliance monitoring requirements
C. Automatic removal of all bias from model predictions
D. Enhanced ability to identify bias and improve model governance
Show Answer
Correct Answer: D
Explanation: Increasing model transparency makes model behavior more interpretable, which helps stakeholders detect biases, understand decisions, and strengthen oversight and governance. It does not eliminate bias, remove regulatory needs, or reduce validation requirements.
$19
Get all 397 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.
