AIF-C01 Fundamentals of Generative AI — 100 Questions

Question 1

A company is using Amazon SageMaker JumpStart to deploy a pre-trained text generation model. After deployment, the model produces slow inference responses. Which action is most likely to improve inference latency?

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

Accepted Answer

Deploy the model on a more powerful instance type with higher GPU memory.. Deploying the model on a more powerful instance type with higher GPU memory directly addresses the computational bottleneck causing slow inference. A larger GPU provides more CUDA cores and memory bandwidth, enabling faster matrix operations and reducing the time per forward pass for the pre-trained text generation model.

Answer

Quantize the model weights to FP16 or INT8.

Answer

Fine-tune the model on a smaller dataset.

Answer

Increase the batch size for inference requests.

Question 2

A company is building a chatbot using Amazon Bedrock. They want to ensure the model's responses are grounded in their internal knowledge base and avoid generating information outside that scope. Which feature should they use?

Accepted Answer

Amazon Bedrock Knowledge Bases. Amazon Bedrock Knowledge Bases is the correct feature because it allows you to connect a foundation model (FM) to your internal data sources, such as documents or databases, and use Retrieval Augmented Generation (RAG) to ground responses in that specific knowledge. This ensures the chatbot only generates information from the provided knowledge base, preventing hallucinations or out-of-scope content.

Answer

Agents for Amazon Bedrock

Answer

Model Evaluation on Amazon Bedrock

Answer

Guardrails for Amazon Bedrock

Question 3

A company is using Amazon Bedrock to generate creative marketing copy. They want to reduce the randomness of the output while maintaining diversity. Which TWO parameters should they adjust?

Accepted Answer

Decrease the top_p value. Option E (Decrease the temperature) is correct because temperature scales the randomness of token sampling in Amazon Bedrock; lowering it makes the probability distribution sharper, so the model favors high-probability tokens and produces more deterministic, less random output. Option D (Decrease the top_p value) is correct because top_p (nucleus sampling) restricts generation to the smallest set of tokens whose cumulative probability exceeds p; reducing p narrows that candidate pool, cutting off low-probability tokens and reducing randomness while still allowing varied word choices within the nucleus, which preserves diversity. Option A (Increase the temperature) is wrong because raising temperature flattens the distribution and increases randomness, the opposite of the goal. Option B (Increase the max token count) is wrong because max tokens only caps output length and does not affect sampling randomness. Option C (Increase the top_k value) is wrong because a larger top_k widens the candidate token set, increasing rather than reducing randomness.

Answer

Increase the temperature

Answer

Increase the max token count

Answer

Increase the top_k value

Question 4

A developer attached this IAM policy to a role used by an application that invokes Claude v2 in us-east-1. The application receives an access denied error. What is the MOST likely cause?

Accepted Answer

The Deny statement is blocking requests because the condition does not match the resource ARN's region. The Deny statement uses a `StringNotEquals` condition on `aws:RequestedRegion` set to `us-east-1`. This means the Deny applies to any request where the requested region is NOT `us-east-1`. Since the resource ARN in the Deny statement is `arn:aws:bedrock:us-east-1::foundation-model/anthropic.claude-v2`, the condition does not match the resource's region (the resource ARN itself is in us-east-1), but the Deny is triggered when the request is made to a different region, blocking the call. The application is likely invoking the model from a region other than us-east-1, causing the Deny to take effect.

Answer

The Allow statement does not include a condition on the region

Answer

The Deny statement uses StringNotEquals instead of StringEquals

Answer

The resource ARN in the Allow statement is incorrect

Question 5

A financial services company is deploying a generative AI model on Amazon SageMaker for real-time fraud detection. The model, a fine-tuned Llama 2 7B, must respond to transaction requests within 500 milliseconds. The team has deployed the model using a SageMaker real-time endpoint with a single ml.g5.2xlarge instance. During load testing, the endpoint achieves an average latency of 450 ms at 10 requests per second (RPS), but the latency spikes to over 2 seconds at 20 RPS. The team needs to maintain sub-500 ms latency at up to 50 RPS. The model is too large to fit on a single GPU, so they are using CPU instances. They considered using a larger instance type but want to minimize cost. What should the team do to meet the latency requirement cost-effectively?

Accepted Answer

Use a SageMaker multi-model endpoint with multiple ml.g5.xlarge instances and auto scaling. A SageMaker multi-model endpoint (MME) allows multiple model replicas to be hosted on a fleet of instances, enabling horizontal scaling to handle increased throughput. By using multiple ml.g5.xlarge instances with auto scaling, the team can distribute the 50 RPS load across several instances, keeping per-instance latency low while minimizing cost compared to a single larger instance. This approach also leverages the fact that the model is too large for a single GPU but can be efficiently served on CPU instances with proper load distribution.

Answer

Upgrade to a single ml.g5.4xlarge instance

Answer

Attach an Amazon Elastic Inference accelerator to the existing instance

Answer

Use SageMaker Serverless Inference to automatically scale

Question 6

Which TWO factors are most important when selecting a foundation model for a sentiment analysis task? (Choose 2)

Accepted Answer

Composition of the model's training data (domain, language). Option A is correct because the composition of the model's training data—its domain and language coverage—directly determines how well the model's learned representations align with the vocabulary, tone, and idioms of the sentiment-bearing text you will feed it, so a model pretrained on matching domains and languages will yield far more accurate sentiment classifications. Option D is correct because model size (number of parameters) is a primary driver of the model's capacity to capture nuanced sentiment, sarcasm, and contextual polarity; larger models generally achieve higher accuracy on complex sentiment tasks, though they demand more compute. Option B is not a primary selection factor because API pricing per invocation is a cost/operational concern, not a determinant of whether the model can perform sentiment analysis well. Option C is not primary because inference latency affects throughput and user experience but does not govern the quality of sentiment predictions. Option E is irrelevant since the color scheme of documentation has no bearing on model capability or task performance.

Answer

API pricing per invocation

Answer

Inference latency

Answer

Color scheme of the model's documentation

Question 7

A user has this IAM policy and attempts to invoke the model in the us-west-2 region. They receive an AccessDenied error. What is the reason?

Accepted Answer

The policy does not allow the us-west-2 region. The policy explicitly denies access to the 'us-west-2' region via the 'Deny' effect with a condition on 'aws:RequestedRegion'. Since the user is attempting to invoke the model in us-west-2, the request matches the deny condition, resulting in an AccessDenied error. The policy does not grant any explicit allow for that region, so the default implicit deny combined with the explicit deny blocks the action.

Answer

The resource ARN is malformed

Answer

The action 'bedrock:InvokeModel' is incorrect

Answer

The model ID is case-sensitive

Question 8

A company is building a customer service chatbot using Amazon Bedrock. Which component of a foundation model determines the creativity and randomness of the generated responses?

Accepted Answer

Temperature. In Amazon Bedrock, the temperature parameter controls the creativity and randomness of a foundation model's generated responses. A higher temperature (e.g., 1.0) makes the output more diverse and creative, while a lower temperature (e.g., 0.2) makes it more deterministic and focused. This directly influences the randomness of token selection.

Answer

Prompt template

Answer

Max tokens

Answer

Top-p

Question 9

A data scientist is deploying a fine-tuned Mistral model on Amazon Bedrock. After deployment, inference latency is too high for real-time applications. Which configuration change can reduce latency without significantly impacting output quality?

Accepted Answer

Reduce max tokens from 1024 to 256. Reducing max tokens from 1024 to 256 directly decreases the number of tokens the model generates per request, which lowers the computational load and memory usage during inference. Since the model only processes up to the specified token limit, this reduces latency proportionally while preserving output quality for short-form responses, as the model's core reasoning and generation capabilities remain unchanged.

Answer

Switch to a larger model variant

Answer

Decrease the top-p to 0.5

Answer

Increase the temperature to 0.9

Question 10

A company fine-tunes a foundation model using SageMaker to create a domain-specific chatbot. After deployment on Bedrock, the model shows high confidence in incorrect answers. What is the most likely cause and its solution?

Accepted Answer

The model is overfitting; apply regularization techniques during fine-tuning. When a fine-tuned model exhibits high confidence in incorrect answers, it is a classic symptom of overfitting, where the model has memorized the training data too closely and fails to generalize to new inputs. Applying regularization techniques such as dropout, weight decay, or early stopping during fine-tuning reduces overfitting by penalizing overly complex decision boundaries, forcing the model to learn more robust patterns. This is the most direct cause and solution given the scenario of a domain-specific chatbot deployed on Bedrock.

Answer

The model was not pre-trained on enough data; use a larger base model

Answer

The training data was imbalanced; collect more diverse data

Answer

The inference temperature is too low; increase it

Question 11

A healthcare startup is using Amazon Bedrock to generate clinical notes. They must prevent the model from outputting any personally identifiable information (PII) such as patient names. What is the most effective approach?

Accepted Answer

Configure a guardrail in Amazon Bedrock to deny PII topics. Amazon Bedrock Guardrails provide a native, policy-based mechanism to deny the generation of PII topics at inference time, without requiring model retraining or external filtering. This approach directly intercepts and blocks prohibited content before it is returned, offering the most reliable and maintainable solution for compliance with healthcare privacy regulations like HIPAA.

Answer

Fine-tune the model on de-identified data only

Answer

Use a prompt engineering technique to instruct the model to avoid PII

Answer

Post-process the output with a regex filter

Question 12

An organization is evaluating different foundation models (FMs) on Amazon Bedrock for a legal document analysis task. Which THREE factors should they consider when selecting a model? (Choose 3.)

Accepted Answer

Model size (number of parameters). Model size (B) matters because the number of parameters strongly influences the model's reasoning capability, accuracy, and latency on complex legal text, so larger models may be needed for nuanced analysis. Support for the specific language of the documents (D) is essential because a model must be trained or fine-tuned on the document's language to correctly interpret legal terminology and produce valid results. Token limits for input and output (E) are critical since legal documents are often long, and the model must accommodate the required context window and response length within Amazon Bedrock's constraints. The region where the model is hosted (A) is not a model-selection factor for task suitability, as it concerns deployment and data-residency rather than the model's analytical capabilities. Cost per inference call (C) is a business consideration but is not one of the three factors specifically tied to matching a model to the legal document analysis task's technical requirements.

Answer

The region where the model is hosted

Answer

Cost per inference call

Question 13

A developer invoked an Amazon Bedrock model and received this output. What does the stopReason field indicate?

Accepted Answer

The model reached the maximum token limit set in the request. The `stopReason` field in Amazon Bedrock model responses indicates why the model stopped generating tokens. When set to `end_turn` or `stop_sequence`, it means the model reached a natural stopping point or encountered a user-defined stop sequence. However, if the value is `max_tokens`, it explicitly means the model stopped because it reached the maximum token limit specified in the `max_tokens` request parameter. Option C correctly identifies this behavior.

Answer

A content filter blocked the output

Answer

The input prompt was too long

Answer

The model reached a natural stopping point

Question 14

A data scientist is unable to invoke the Claude v2 model from an EC2 instance with IP 10.0.1.5. What is the most likely reason?

Accepted Answer

The resource ARN is incorrect because it should not include the account ID. The most likely reason is that the resource ARN in the IAM policy is incorrect because it should not include the account ID. For Bedrock model invocation, the resource ARN format is `arn:aws:bedrock:region::foundation-model/model-id`, where the account ID is omitted (replaced by a double colon). Including the account ID (e.g., `arn:aws:bedrock:region:account-id:foundation-model/model-id`) makes the ARN invalid, causing the policy to fail and denying the invoke action from the EC2 instance.

Answer

The condition is too restrictive

Answer

The policy does not allow the ec2:AssociateAddress action

Answer

The model is not available in us-east-1

Question 15

A company wants to use Amazon Bedrock to generate images from text descriptions. Which model should they use?

Accepted Answer

Amazon Titan Image Generator. Amazon Titan Image Generator is the correct choice because it is specifically designed by AWS for generating images from text descriptions within Amazon Bedrock. It leverages a diffusion-based architecture to create high-quality images, making it the native and optimal model for this task on the platform.

Answer

Stable Diffusion XL

Answer

Amazon Titan Text

Answer

Amazon Polly

Question 16

A research team needs to generate high-quality images with Amazon Bedrock that are realistic and consistent with a specific artistic style. Which combination of parameters should they use?

Accepted Answer

Use a CFG (classifier-free guidance) scale and include a style prompt. Classifier-free guidance (CFG) scale controls how strongly the model adheres to the prompt, and including a style prompt (e.g., 'in the style of Impressionism') directly conditions the generation to produce images consistent with a specific artistic style. This combination allows the research team to balance prompt fidelity (via CFG scale) and stylistic control (via the style prompt), which is essential for generating realistic, style-consistent images in Amazon Bedrock.

Answer

High temperature, low top_p

Answer

Low temperature, high top_p

Answer

Increase the number of steps and reduce the number of samples

Question 17

A company wants to automate the extraction of key information from customer support tickets using generative AI. They have a small labeled dataset. Which approach would be most effective?

Accepted Answer

Fine-tune a foundation model on the labeled data. Fine-tuning a foundation model on a small labeled dataset is the most effective approach because it adapts the pre-trained model's knowledge to the specific task of extracting key information from support tickets, requiring far less data and compute than training from scratch. This leverages transfer learning, where the model already understands language structure and can learn the nuances of the target extraction task with minimal additional examples.

Answer

Use zero-shot prompting with a foundation model

Answer

Train a custom model from scratch

Answer

Use Amazon Comprehend custom entity recognition

Question 18

Which THREE factors should be considered when selecting a foundation model for a text generation task? (Select three.)

Accepted Answer

Context window length. Option A (Context window length) is correct because the context window determines how many tokens the model can accept as input plus generate as output in a single request, which directly constrains whether long prompts, documents, or multi-turn conversations can be handled for the text generation task. Option C (Model license) is correct because the license governs whether the model may be used commercially, modified, or redistributed, and it dictates legal and cost implications for deploying the model in a production text generation workload. Option D (Number of parameters) is correct because parameter count is a primary driver of the model's capability, quality, and resource requirements, so it must be matched to the task's accuracy needs and the available compute budget. Option B (Inference latency) is not among the marked correct answers here, and Option E (AWS Region availability) is likewise not marked, so neither should be treated as a required selection factor for this question.

Answer

Inference latency

Answer

AWS Region availability

Question 19

Refer to the exhibit. A user invoked a Claude model using provisioned throughput and received a ThrottlingException. Which is the most likely cause?

Accepted Answer

The provisioned throughput request per minute limit was exceeded. Provisioned throughput guarantees a specific number of processing units (TPUs) for your model invocation. When you exceed the requests-per-minute (RPM) or tokens-per-minute (TPM) limit allocated to your provisioned throughput, Amazon Bedrock returns a ThrottlingException. Option B correctly identifies that the user hit the provisioned throughput request per minute limit.

Answer

The model is not available in the region

Answer

The prompt was too long

Answer

The inference type should be ON_DEMAND

Question 20

A company is building a customer support chatbot using Amazon Bedrock. They have a large corpus of internal documentation and want to provide accurate answers without retraining the model. Which approach should they use?

Accepted Answer

Use Retrieval Augmented Generation (RAG) with Amazon Bedrock Knowledge Base.. Retrieval Augmented Generation (RAG) with Amazon Bedrock Knowledge Base is the correct approach because it allows the chatbot to retrieve relevant chunks from the internal documentation in real time and pass them as context to the foundation model. This provides accurate, up-to-date answers without retraining, directly addressing the requirement to avoid model customization.

Answer

Use advanced prompt engineering with a generic foundation model.

Answer

Use a pre-trained foundation model without customization.

Answer

Fine-tune the model on their documentation using Amazon SageMaker.