AIF-C01 Practice Test 26 — 15 Questions

Question 1

A data scientist is comparing two foundation models for a text classification task. They want to use automated metrics to evaluate performance on a labelled test set. Which metric is most appropriate for a multi-class classification problem?

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

Accepted Answer

F1 score. F1 score is a standard metric for classification, balancing precision and recall. ROUGE and BLEU are for text generation, Perplexity for language modelling.

Answer

ROUGE-L

Answer

BLEU

Answer

Perplexity

Question 2

A financial services company uses a foundation model for document analysis. They need to ensure the model does not output sensitive customer information from its training data. What is the most effective mitigation?

Accepted Answer

Implement output filtering using an external service. Output filtering using an external service is the most effective mitigation because it acts as a post-processing layer that can detect and redact sensitive customer information (e.g., PII, account numbers) before the model's response is returned to the user. This approach does not rely on the model's internal training or input modifications, which can be bypassed or incomplete. It provides a robust, policy-driven control that can be updated independently of the model.

Answer

Choose a model that has been fine-tuned on financial data

Answer

Apply data masking before sending input

Answer

Use a private endpoint

Question 3

A company runs a question-answering application on Amazon Bedrock that answers from a large knowledge base. Recently, users have reported that the model gives incomplete answers, often missing details from the middle of documents. The team suspects the chunking strategy is suboptimal. Which adjustment is MOST likely to improve completeness?

Accepted Answer

Use overlapping chunks so that boundaries do not cut off meaningful content. Overlapping chunks ensure that sentences or concepts that span chunk boundaries are not lost, improving retrieval completeness. Larger chunks may cause loss of precision, smaller chunks may lose context, and embedding model change is not directly related to chunking.

Answer

Change the embedding model to a more powerful one

Answer

Increase the chunk size to include more context per chunk

Answer

Decrease the chunk size to capture more granular details

Question 4

A financial institution uses a machine learning model to approve loan applications. The model is trained on historical data that includes biased lending practices. What is the most effective first step to address potential bias?

Accepted Answer

Audit the training data for bias and review feature selection. The most effective first step to address potential bias is to audit the training data for bias and review feature selection. This aligns with the AWS Responsible AI guidelines, which emphasize that bias mitigation must start with understanding the data and features before any model changes. Without auditing, you cannot identify the source of bias—whether it's in the labels, sampling, or feature correlations—making subsequent steps like retraining or deployment ineffective or harmful.

Answer

Immediately deploy the model and monitor for biased outcomes

Answer

Retrain the model with synthetic data generated from the original dataset

Answer

Remove all demographic features from the model

Question 5

A data scientist is building a text classification system using Amazon Bedrock. They want to evaluate different foundation models for accuracy and latency. Which TWO approaches are appropriate for comparing models? (Select TWO.)

Accepted Answer

Invoke several models with a sample set of queries and compare the outputs manually. Option B is correct because invoking several models through the Amazon Bedrock InvokeModel API with the same sample set of queries lets the data scientist directly observe and compare real outputs, including quality and latency, under identical conditions. Option D is correct because creating a labeled test dataset with ground truth and computing accuracy metrics (e.g., precision, recall, F1) provides an objective, quantitative evaluation of each foundation model's classification performance. Option A is not appropriate because building a custom ensemble model from scratch is a separate modeling effort, not a method for comparing existing Bedrock foundation models. Option C is not appropriate because parameter count from documentation does not reliably predict task-specific accuracy or latency, and no empirical testing is performed. Option E is not appropriate because choosing the lowest cost per token without testing ignores accuracy and latency requirements, which are the stated evaluation criteria.

Answer

Build a custom model from scratch that combines the outputs of all models

Answer

Read the model documentation and assume the one with the largest parameter count is best

Answer

Select the model with the lowest cost per token without testing

Question 6

A company is developing a fraud detection system using a neural network. The training loss decreases steadily but the validation loss begins to increase after a certain number of epochs. Which action should be taken to address this issue?

Accepted Answer

Implement early stopping to halt training when validation performance degrades. The described behavior—training loss decreasing while validation loss increases—is a classic sign of overfitting. Early stopping monitors the validation loss and halts training when it stops improving (or begins to degrade), preventing the model from memorizing noise in the training data. This directly addresses the overfitting issue without requiring architectural or data changes.

Answer

Add more layers to the neural network

Answer

Increase the amount of training data

Answer

Increase the learning rate

Question 7

A company is building a generative AI application for code generation. They want to minimize costs while maintaining acceptable performance for their workload, which has periodic spikes in demand. Which approach would be MOST cost-effective?

Accepted Answer

Right-size model selection: use a smaller model for simple tasks and a larger model only when needed. Right-sizing model selection involves using smaller, cheaper models for simple tasks and reserving larger, more expensive models for complex tasks. This optimizes cost while maintaining acceptable performance, especially for workloads with periodic spikes. It avoids over-provisioning and leverages the most cost-effective model for each request.

Answer

Always use the largest available foundation model for all requests

Answer

Use model caching to store responses for repeated prompts

Answer

Use batch inference for all requests

Question 8

A developer is using the Amazon Bedrock API to generate text. They notice that the model sometimes returns harmful content despite setting safety parameters. What is the BEST way to add an additional layer of content filtering?

Accepted Answer

Configure content filters in Amazon Bedrock Guardrails. Amazon Bedrock Guardrails provides a dedicated, configurable content filtering layer that can block harmful content at inference time, independent of the model's built-in safety parameters. This allows developers to enforce custom policies (e.g., hate speech, violence) without modifying the model itself, making it the best additional safeguard.

Answer

Fine-tune the model on a curated safe dataset

Answer

Improve prompt engineering with more specific instructions

Answer

Use AWS WAF to filter API responses

Question 9

A financial services firm fine-tuned a generative AI model on Amazon SageMaker to summarize quarterly reports. The summaries often miss key financial metrics such as revenue and profit margins. The fine-tuning dataset contained full reports with summaries that included these metrics. The model appears to understand the reports but omits critical numbers. Which course of action would most likely improve the summaries?

Accepted Answer

Re-fine-tune using a carefully crafted dataset that includes explicit instructions to include key metrics and provides examples of correct summaries. The model's failure to include key financial metrics despite having them in the training data indicates a misalignment between the training objective and the desired output. By re-fine-tuning with a dataset that explicitly instructs the model to include key metrics and provides correct examples, you directly teach the model to prioritize and extract those specific numerical values during summarization, addressing the root cause of omission.

Answer

Increase the maximum number of tokens in the summary

Answer

Switch to a different pre-trained model like Claude instead of the current one

Answer

Implement a post-processing Lambda function that extracts metrics from the original report and appends them to the summary

Question 10

Which TWO techniques are commonly used to prevent overfitting in machine learning models? (Select TWO.)

Accepted Answer

Use cross-validation. Option B (Use cross-validation) is correct because techniques like k-fold cross-validation estimate how well a model generalizes to unseen data by training and validating on different data splits, which helps detect and mitigate overfitting during model selection and hyperparameter tuning. Option E (Use regularization) is correct because methods such as L1 (Lasso) and L2 (Ridge) regularization add a penalty term to the loss function that constrains the magnitude of model weights, reducing variance and preventing the model from fitting noise in the training data. The unmarked options do not belong: adding more irrelevant features (A) increases dimensionality and noise, making overfitting worse; increasing model complexity (C) raises variance and typically worsens overfitting; and reducing the amount of training data (D) gives the model less information to learn generalizable patterns, also promoting overfitting.

Answer

Add more irrelevant features

Answer

Increase model complexity

Answer

Reduce the amount of training data

Question 11

A company operates in a region where Amazon Bedrock is not available. They want to use generative AI but must keep data within the country. Which solution should they consider?

Accepted Answer

Use Amazon SageMaker to host an open-source model in the local region.. Amazon SageMaker allows you to host open-source models (e.g., Llama 2, Falcon) in any AWS region, including those where Bedrock is unavailable. This satisfies the data residency requirement because the model and data never leave the local region. SageMaker provides full control over the infrastructure, enabling compliance with local data sovereignty laws.

Answer

Wait for Bedrock to become available in their region; there is no alternative.

Answer

Use Amazon Bedrock in the nearest available region with cross-region inference.

Answer

Use an API from a third-party generative AI provider with AWS PrivateLink.

Question 12

A bank wants to use Amazon Augmented AI (A2I) to review high-value loan applications that require human judgment. Which workflow best implements human-in-the-loop review for these predictions?

Accepted Answer

Set up an A2I workflow with a confidence threshold, so low-confidence predictions are sent to human reviewers. Amazon A2I enables human review of low-confidence predictions or specific conditions. The best practice is to set a confidence threshold; predictions below that threshold are sent to human reviewers.

Answer

Route all loan applications to human reviewers for approval

Answer

Use Amazon Mechanical Turk to review all predictions in real time

Answer

Configure a private workforce in Amazon SageMaker Ground Truth

Question 13

A data scientist is preparing data for a classification task. Which TWO techniques are commonly used for handling missing values? (Choose two.)

Accepted Answer

Imputing with mean. Replacing missing numeric entries with the column mean is a standard, simple imputation technique that preserves the dataset size and avoids discarding information. Option D (Dropping rows with any missing values) is also correct because listwise deletion is a common, straightforward approach to handling missing data, especially when the missingness is minimal or random. The other options do not address missing values: A (Label encoding) converts categorical labels into integer codes, B (Normalization) rescales numeric feature ranges, and E (One-hot encoding) creates binary indicator columns for categories—all are preprocessing steps for existing values, not missing-data handling.

Answer

Label encoding

Answer

Normalization

Answer

One-hot encoding

Question 14

A financial institution is building a model to approve loan applications. They must comply with the EU AI Act, which classifies credit scoring as a high-risk AI system. Which requirement is the MOST likely to apply under the EU AI Act?

Accepted Answer

The system must be able to be overridden or stopped by a human. Under the EU AI Act, high-risk AI systems (including credit scoring) must ensure human oversight, meaning the system can be overridden or stopped by a human. This is a core requirement in Article 14, ensuring meaningful human control over high-risk decisions. Other requirements include risk management, data governance, and transparency, but human oversight is explicitly mandated.

Answer

The model must be trained exclusively on data from within the EU

Answer

The model must achieve a minimum accuracy of 95%

Answer

The system must be explainable using SHAP values

Question 15

A company is building a RAG application with Amazon Bedrock Knowledge Bases. They want to ensure that the retriever returns the most semantically relevant chunks. They are using a large document corpus with many similar passages. Which chunking strategy is MOST likely to improve retrieval accuracy?

Accepted Answer

Semantic chunking that splits at paragraph or section boundaries. Semantic chunking splits documents at natural boundaries like paragraphs or sections, preserving the coherence of each chunk. This ensures that each chunk contains a complete, self-contained idea, which allows the retriever to match the semantic meaning of the query more accurately. In a corpus with many similar passages, this approach reduces noise and improves the relevance of retrieved chunks.

Answer

Fixed‑size token chunking with no overlap

Answer

Overlapping fixed‑size chunks with 50% overlap

Answer

Very small chunks (50 tokens) to maximize granularity