AIF-C01 Practice Test 5 — 25 Questions

Question 1

Which TWO are best practices for model monitoring in production on AWS?

Accepted Answer

Monitor input data drift. Option C (Monitor input data drift) is correct because production models degrade when the statistical distribution of incoming features diverges from training data, so tracking data drift with tools like SageMaker Model Monitor detects this covariate shift early. Option E (Monitor prediction drift) is correct because shifts in the distribution of model outputs (e.g., changing class proportions or score distributions) signal concept drift or upstream data issues that require investigation or retraining. Together, input and prediction drift monitoring are the two core best practices for maintaining model quality in production. Option A is wrong because disabling logging removes the observability needed to detect drift, errors, and bias, and logging overhead is typically negligible relative to inference. Option B is wrong because instance type (CPU vs. GPU) is a performance/cost choice, not a monitoring best practice. Option D is wrong because retraining daily is not a best practice by itself; retraining should be triggered by monitored drift or performance degradation, and daily retraining can be costly and destabilizing.

Answer

Disable logging to reduce latency

Answer

Use only CPU instances

Answer

Retrain model daily

Question 2

A company needs to convert a large number of recorded customer service calls into text for analysis. Which AWS service should they use?

Accepted Answer

Amazon Transcribe. Amazon Transcribe is the correct service because it is specifically designed to convert speech from audio files (such as recorded customer service calls) into text using automatic speech recognition (ASR). This directly matches the requirement to transcribe audio content for subsequent text analysis.

Answer

Amazon Textract

Answer

Amazon Comprehend

Answer

Amazon Polly

Question 3

A data scientist is using a foundation model to summarize long documents. Which TWO of the following steps are most likely to improve the quality of the summaries?

Accepted Answer

Break the input document into chunks and summarize each chunk separately.. Option A is correct because chunking a long document and summarizing each chunk separately keeps each request within the model's effective context window, avoiding truncation and the degraded recall that occurs when a foundation model must attend to very long inputs, and the chunk summaries can then be combined into a final summary. Option C is correct because few-shot prompting supplies concrete examples of the desired summary style, length, and level of detail, which steers the model's output distribution toward the target format more reliably than a bare instruction. Option B is not appropriate because a high temperature increases randomness and creativity, which harms factual fidelity and consistency in summarization. Option D is not the priority here because frequency penalty only discourages repeated tokens and does not address the core problems of long-input context limits or output style alignment. Option E is not correct because increasing max tokens only raises the output length cap; it does not extend the model's usable context window for the input document and can even encourage overly long, unfocused summaries.

Answer

Use a high temperature parameter to increase creativity.

Answer

Use a low frequency penalty to reduce repetition.

Answer

Use a longer context length by increasing the max tokens parameter.

Question 4

A data science team is using Amazon SageMaker to train multiple models with different hyperparameters. They want to track metrics, compare runs, and reproduce the best result. Which SageMaker feature should they use?

Accepted Answer

SageMaker Experiments. SageMaker Experiments is the correct feature because it is specifically designed to track, organize, and compare machine learning training runs (trials) with different hyperparameters and metrics. It allows data scientists to log parameters, metrics, and artifacts for each run, compare results across runs, and retrieve the exact configuration needed to reproduce the best-performing model.

Answer

SageMaker Model Registry

Answer

SageMaker Debugger

Answer

SageMaker Autopilot

Question 5

Refer to the exhibit. A data scientist ran a training job on Amazon SageMaker. The job failed with the error shown. What is the most likely cause?

Accepted Answer

The batch size is too large for the instance's GPU memory. The error message indicates a CUDA out-of-memory error, which occurs when the GPU memory is insufficient for the requested batch size. Option D is correct because increasing the batch size beyond the GPU's memory capacity causes the training job to fail with this specific error.

Answer

The S3 input path is incorrect

Answer

The IAM role does not have permission to access S3

Answer

The training code has a syntax error

Question 6

A company is deploying a generative AI application for customer support. They need to ensure that the model does not generate responses containing personally identifiable information (PII) even if it appears in the retrieved context. Which Bedrock feature should they configure?

Accepted Answer

Bedrock Guardrails. Bedrock Guardrails include PII detection and redaction. Knowledge Bases retrieve context but do not filter PII. Agents orchestrate tasks but do not enforce PII rules. Model evaluation tests performance but does not provide runtime controls.

Answer

Bedrock Knowledge Bases

Answer

Bedrock Agents

Answer

Bedrock Model Evaluation

Question 7

An ML team is deploying a real-time inference endpoint for a computer vision model using Amazon SageMaker. The model requires GPU acceleration for low latency. Which instance type should the team choose to minimize cost while meeting the GPU requirement?

Accepted Answer

ml.g5.xlarge. (ml.g5.xlarge) is correct because it provides a GPU (NVIDIA A10G Tensor Core GPU) necessary for low-latency GPU acceleration in computer vision inference, while being the most cost-effective GPU instance among the options. The ml.g5.xlarge offers sufficient GPU compute for real-time inference at a lower hourly cost compared to ml.p3.2xlarge, making it the optimal choice for minimizing cost while meeting the GPU requirement.

Answer

ml.c5.xlarge

Answer

ml.p3.2xlarge

Answer

ml.p4d.24xlarge

Question 8

A data scientist is using Amazon SageMaker to train a large language model from scratch. Which AWS service is most suitable for managing the training infrastructure, including automatic scaling and spot instance recovery?

Accepted Answer

Amazon SageMaker Training job.. Amazon SageMaker Training jobs are the most suitable service for managing training infrastructure because they provide built-in automatic scaling, managed spot instance recovery, and distributed training orchestration. This allows the data scientist to focus on model development rather than provisioning and managing EC2 instances, load balancers, or recovery scripts.

Answer

AWS Lambda function.

Answer

Amazon SageMaker Notebook instance.

Answer

Amazon EC2 with a custom setup.

Question 9

Which THREE steps are typically involved in fine-tuning a foundation model? (Select THREE.)

Accepted Answer

Prepare a labeled dataset specific to the target domain. Fine-tuning a foundation model begins with selecting an appropriate pre-trained foundation model as the starting point (D), since the whole point of fine-tuning is to adapt existing general-purpose weights rather than train from scratch. Next, you must prepare a labeled dataset specific to the target domain (B), because supervised fine-tuning requires task-relevant input-output pairs to steer the model toward the desired behavior. Then you train the model on that domain dataset with a lower learning rate (C), which is standard practice to avoid catastrophic forgetting and to gently nudge the pre-trained weights instead of overwriting them. Option A is incorrect because deploying the model immediately without additional training is the opposite of fine-tuning—it describes using the base model as-is. Option E is incorrect because fine-tuning does not require choosing an architecture with more parameters than the base model; you typically fine-tune the same pre-trained architecture, and increasing parameter count is not a fine-tuning step.

Answer

Deploy the model immediately without additional training

Answer

Choose a model architecture with more parameters than the base model

Question 10

A media company uses a generative AI model to automatically create image captions for user-uploaded photos. During quality assurance, testers discover that the model sometimes generates captions that include stereotypes based on gender and race, even when the photos do not contain people. For example, a photo of a kitchen produces captions like 'woman cooking,' and a photo of a sports car generates 'man driving.' The company wants to launch the feature soon but recognizes the reputational risk. They have a limited budget and need to implement a solution that reduces harmful stereotypes without overly restricting the captions' creativity. The team has access to the model's training data, which is a large public dataset of image-caption pairs. Which approach should the team prioritize?

Accepted Answer

Filter the training data to remove or downweight pairs with stereotypes, then fine-tune the model. The root cause of the stereotypical captions is the training data itself, which contains biased image-caption pairs (e.g., kitchens associated with women, sports cars with men). Filtering or downweighting these biased pairs and then fine-tuning the model directly addresses the source of the bias, reducing harmful stereotypes while preserving the model's generative creativity. This approach is cost-effective because it leverages the existing model and dataset, and it aligns with AWS responsible AI practices of data-centric debiasing. Post-processing or model replacement would not fix the underlying bias and could either over-restrict or fail to generalize.

Answer

Replace the generative model with a simpler classification model that only describes objects

Answer

Use a different pre-trained generative model that is larger and more accurate

Answer

Add a post-processing filter that checks captions for known stereotype patterns and blocks them

Question 11

A company uses Amazon Bedrock to generate content. They need to guarantee that the model does not produce toxic or biased outputs and that any violations are logged for review. Which service should they integrate with Bedrock?

Accepted Answer

Amazon Bedrock Guardrails. Amazon Bedrock Guardrails is the correct service because it is specifically designed to enforce content policies, filter toxic or biased outputs, and log violations for review within Bedrock workflows. It provides configurable thresholds for harmful categories (e.g., hate, insults, sexual content) and integrates directly with Bedrock models to intercept and redact or block undesired responses, with all violations recorded in Amazon CloudWatch Logs.

Answer

Amazon Comprehend

Answer

AWS WAF

Answer

AWS CloudTrail

Question 12

A retail company uses a machine learning model to forecast daily product demand. The model is a time series model that uses historical sales data. The model has been performing well, but recently the forecasts have been consistently too low, leading to stockouts. The data scientist notices that the model was trained on data up to last year, and the company has since launched a successful marketing campaign that increased sales by 20%. The data scientist needs to update the model to reflect the new sales patterns. Which approach should the data scientist take?

Accepted Answer

Retrain the model on the most recent data that includes the sales from the marketing campaign.. Retraining the model on the most recent data that includes the sales from the marketing campaign allows the model to learn the new underlying pattern in the time series. Since the model is a time series model, it relies on historical patterns to make forecasts; retraining on data that captures the 20% sales lift ensures the model adapts to the new demand level, reducing the persistent underforecasting.

Answer

Add a feature for the marketing campaign and continue using the old model.

Answer

Switch to a different model type, such as ARIMA, without retraining.

Answer

Multiply the model's predictions by 1.2 to account for the marketing campaign.

Question 13

A team is evaluating two generative AI models for a summarization task. They have reference summaries and generated summaries. Which automated metric is BEST suited to measure the overlap of n-grams between the generated and reference summaries?

Accepted Answer

ROUGE. ROUGE (Recall-Oriented Understudy for Gisting Evaluation) measures n-gram overlap between generated and reference text, and is commonly used for summarization. BLEU is for translation, BERTScore uses embeddings, and perplexity measures language model confidence.

Answer

BLEU

Answer

BERTScore

Answer

Perplexity

Question 14

A data scientist is setting up a RAG pipeline using Amazon Bedrock Knowledge Bases. They need to ingest documents, split them into chunks, generate embeddings, and store them for retrieval. Which TWO steps are part of the ingestion process? (Choose TWO)

Accepted Answer

Parse the documents and split them into chunks. Option D is correct because the ingestion (data source sync) process in Amazon Bedrock Knowledge Bases begins by parsing the source documents and splitting them into manageable chunks during the chunking step, which is required before embeddings can be created. Option E is correct because after chunking, the service invokes a selected Amazon Bedrock embedding model (for example, Amazon Titan Embeddings or Cohere Embed) to convert each chunk into a vector representation that is then written to the vector store. Option A is not part of the standard ingestion flow; while Lambda functions can be used with custom transformations or orchestration, defining a Lambda for post-processing is not a required ingestion step. Option B is incorrect because similarity search is a retrieval-time operation performed against the vector store, not part of ingestion. Option C is also incorrect because augmenting the user prompt with retrieved chunks happens during the retrieval/generation (query) phase, after ingestion is complete.

Answer

Define a Lambda function for post-processing

Answer

Perform a similarity search on the vector store

Answer

Augment the user prompt with retrieved chunks

Question 15

Which TWO techniques can reduce the cost of running a fine-tuned foundation model on Amazon SageMaker? (Choose TWO.)

Accepted Answer

Implement structured pruning to remove less important model parameters.. Option A is correct because structured pruning removes less important weights, neurons, or channels from the fine-tuned model, producing a smaller model that requires fewer compute and memory resources during SageMaker inference, which directly lowers hosting cost. Option C is correct because quantization reduces numerical precision from FP32 to FP16 or INT8, shrinking model size and memory bandwidth needs and enabling faster, cheaper inference on SageMaker endpoints, especially with GPU instances that support lower-precision arithmetic. Option B is not correct because using larger GPU instances increases the hourly cost of the endpoint rather than reducing it. Option D is not correct because keeping parameters in FP32 preserves accuracy but consumes more memory and compute, raising cost. Option E is not correct because increasing training epochs affects training time and accuracy, not the cost of running inference on the deployed model.

Answer

Use larger instance types with more GPUs to speed up inference.

Answer

Store the model parameters in FP32 to maintain accuracy during inference.

Answer

Increase the number of training epochs to achieve higher accuracy.

Question 16

A data governance team wants to enforce fine-grained access control on data in an Amazon S3 data lake used by multiple business units for AI training. Which AWS service should they use to define and manage data permissions at the table and column level?

Accepted Answer

AWS Lake Formation. AWS Lake Formation is purpose-built to define and manage fine-grained access control (table, column, row, and cell level) on data stored in Amazon S3 data lakes. It provides a centralized permission model that works across analytics and AI/ML services, making it the correct choice for multi-business-unit governance. IAM alone cannot express column-level permissions on S3 data.

Answer

AWS Identity and Access Management (IAM)

Answer

AWS CloudTrail

Answer

Amazon Macie

Question 17

A company has deployed a model on Amazon SageMaker and enabled Model Monitor. They notice that the model's prediction accuracy has declined over time. Which type of drift is this, and what should they do?

Accepted Answer

This is concept drift; they should retrain the model with new data. The decline in prediction accuracy over time, despite the model and data pipeline remaining the same, is characteristic of concept drift—the relationship between input features and the target variable has changed. The appropriate action is to retrain the model with new data that reflects the current relationship. SageMaker Model Monitor can detect this via accuracy metrics.

Answer

This is bias drift; they should run SageMaker Clarify

Answer

This is model drift; they should delete and redeploy the model

Answer

This is data drift; they should update the training data

Question 18

An e-commerce company is using a foundation model to generate product descriptions. They want to reduce costs by caching frequently requested descriptions. Which AWS service should they use to implement a cache?

Accepted Answer

Amazon ElastiCache. Amazon ElastiCache is the correct choice because it provides an in-memory caching layer (using Redis or Memcached) that can store frequently requested product descriptions, reducing the need to invoke the foundation model repeatedly. This directly lowers inference costs and latency by serving cached responses instead of generating new ones each time.

Answer

Amazon CloudFront

Answer

Amazon DynamoDB

Answer

Amazon S3

Question 19

A company is using Amazon Bedrock to generate marketing content. They want to evaluate the quality of the generated text. Which TWO metrics are most appropriate for evaluating text quality?

Accepted Answer

Perplexity. Perplexity (B) is a standard intrinsic metric for generative language models that measures how well a probability distribution predicts a sample, with lower values indicating the model is less surprised by the text and thus produces more fluent, coherent output. BLEU (E) is a reference-based metric that compares n-gram overlap between generated text and one or more reference texts, making it well suited for evaluating the quality of machine-generated marketing copy against human-written examples. Precision (A), Accuracy (C), and F1 score (D) are classification metrics that require discrete predicted labels and ground-truth classes, so they do not apply to free-form text generation where the output is a sequence of tokens rather than a category.

Answer

Precision

Answer

Accuracy

Answer

F1 score

Question 20

A developer sends the above request to Amazon Bedrock with Anthropic Claude. The model returns a response that stops before reaching 500 tokens. What is the most likely reason?

Accepted Answer

The model reached a stop sequence. The model stopped before reaching 500 tokens because the request likely included a stop sequence (e.g., `

Human:` or a custom stop token) that matched the generated output. When a stop sequence is encountered, Bedrock immediately halts generation, even if the token limit has not been reached. This is the most direct explanation for a premature stop.

Answer

The temperature is set too high

Answer

The model is not trained on this topic

Answer

The token limit is exceeded