Amazon AIP-C01 Exam Dumps

AWS Generative AI Developer Professional

831 Reviews

Exam Code	AIP-C01
Exam Name	AWS Generative AI Developer Professional
Questions	75 Questions Answers With Explanation
Update Date	December 29,2025
Price	Was : $142.2 Today : $79 Was : $160.2 Today : $89 Was : $178.2 Today : $99

Add To Cart

Demo Questions

Question # 1

Question on LCNC LLM Fine-Tuning ? Category: AIP – Operational Efficiency and Optimization for Generative AI Applications. ? Scenario: A team needs to fine-tune an LLM for text summarization using a lowcode/no-code (LCNC) solution to automate model training and minimize manual intervention. ? Question: Which solution will best meet the team’s requirements?

Utilize SageMaker Studio for fine-tuning an LLM deployed on Amazon EC2 instances, simplifying the training process with an interactive and intuitive environment.
Leverage SageMaker Script Mode to fine-tune an LLM on Amazon EC2 instances, enabling custom training scripts to optimize model performance with flexibility
Configure SageMaker Autopilot to fine-tune an LLM deployed via SageMaker JumpStart, streamlining model customization with automatic setup and minimal user intervention.

Show Answer

Question # 2

Question on Cold-Start Forecasting ? Category: AIP – Foundation Model Integration, Data Management, and Compliance. ? Scenario: A manufacturer needs to forecast weekly sales for a brand-new product variant that has no sales history (cold-start problem). The model must learn shared patterns across existing SKUs. ? Question: Which approach best satisfies these requirements?

Use SageMaker AI to train a Linear Learner regression model using historical sales data as features and forecast values as labels for all SKUs. 
Use SageMaker AI to train a Random Cut Forest (RCF) model to detect anomalies in historical sales data and project future demand levels for the new variant
Use SageMaker AI to train a K-means clustering model to group similar SKUs and infer demand patterns for the new variant based on the nearest cluster.
Use SageMaker AI to train the built-in DeepAR algorithm across all related SKUs and then generate a forecast for the new variant. 

Show Answer

Question # 3

. Question on Canvas Access to External Model (Select TWO) ? Category: AIP – Implementation and Integration. ? Scenario: An LLM was fine-tuned outside of SageMaker, and artifacts are in S3. A non-technical team (data specialists) needs access to this model via SageMaker Canvas. ? Question: Which combination of steps must be taken for the AI developer to enable SageMaker Canvas access to the model? (Select TWO.).

 The AI developer is required to set up a SageMaker endpoint for the model
The data specialist team must create a shared workspace within SageMaker Canvas that allows both the AI developer and data specialists to access the model.
The AI developer must convert the model into a TensorFlow or PyTorch format for SageMaker Canvas compatibility
The AI developer must register the model in the SageMaker Model Registry to enable the data specialist team?s access via SageMaker Canvas
The data specialist team must be granted the necessary permissions to access the S3 bucket where the model artifacts are stored

Show Answer

Question # 4

Question on Multi-Dimensional Visualization ? Category: AIP – Operational Efficiency and Optimization for Generative AI Applications. ? Scenario: Visualize recommendation results across four dimensions in SageMaker Canvas: X-axis (interest score), Y-axis (conversion rate), Color (product category), and Size (number of impressions). ? Question: Which approach best satisfies the given requirements?.

Visualize the data using the SageMaker Data Wrangler scatter plot visualization and color data points by the third feature to represent all four dimensions.
Use the SageMaker Canvas Box Plot visualization to compare distributions and use a fill pattern for the third dimension.
Use the SageMaker Canvas Bar Chart visualization to group products by category and simultaneously apply bar color and height to represent interest score and conversion rate
Apply the SageMaker Canvas scatter plot visualization and map the third dimension (product category) to scatter point color and the fourth dimension (number of impressions) to scatter point size

Show Answer

Question # 5

Question on Toxic Language Detection ? Category: AIP – AI Safety, Security, and Governance. ? Scenario: A social media platform needs to enhance safety by detecting toxic or harmful language in real-time (hate speech, harassment) within its SageMaker AI inference pipeline. The solution must be managed, handle high throughput, and provide confidence scores. ? Question: Which of the solutions provides a managed solution for detecting toxicity in text to support this ML pipeline?. 

Use Amazon Bedrock to fine-tune a foundation model for general language understanding
Utilize Amazon Comprehend sentiment analysis to detect negative comments and block content automatically
Use Amazon Translate to convert text into another language before moderation to reduce offensive content
. Utilize Amazon Comprehend toxicity detection to identify abusive or harmful language in text. 

Show Answer

Question # 6

Question on Multilingual Content Processing ? Category: AIP – Implementation and Integration. ? Scenario: A multinational company needs an efficient solution to process audio/video content, translate it from Spanish (and other languages) into English, and summarize it quickly using an LLM, minimizing deployment time and maximizing scalability. ? Question: Which option will best fulfill these requirements in the shortest time possible?.

Train a custom model in Amazon SageMaker AI to process the data into English, then deploy an LLM in SageMaker AI for summarizing the content.
Leverage Amazon Translate to translate the text into English, apply a pre- trained model in Amazon SageMaker AI for analysis, and summarize the content using the Claude Anthropic model in Amazon Bedrock. 
Use AWS Glue to clean and prepare the data, then use Amazon Translate to translate the data into English, and summarize the content using Amazon Lex to create a conversational summary.
. Utilize Amazon Transcribe for audio and video-to-text conversion, Amazon Translate for translating the content into English, and Amazon Bedrock with the Jamba model for summarizing the text.

Show Answer

Question # 7

Question on Predictive Maintenance (LCNC) ? Category: AIP – AI Safety, Security, and Governance. ? Scenario: A company analyzes maintenance reports (Comprehend extraction) and sensor readings (S3) to predict equipment failure. An analytics team must prepare the combined dataset and train a custom predictive model using an interface that simplifies data preparation and model training while maintaining integration with other SageMaker components. ? Question: Which should be used to prepare the data and train a custom model for predicting equipment maintenance needs?.

Utilize Amazon Bedrock to fine-tune a foundation model to predict equipment failures from sensor data and maintenance notes.
Use SageMaker Ground Truth to label maintenance events and automatically train a predictive model for maintenance scheduling. 
Utilize Comprehend to directly build and deploy a predictive model for maintenance events without relying on SageMaker services.
Use SageMaker Canvas to prepare the combined dataset and train a custom model through a no-code interface that integrates with SageMaker AI.

Show Answer

Question # 8

Question on Edge ML Deployment ? Category: AIP – Foundation Model Integration, Data Management, and Compliance. ? Scenario: A manufacturing company in remote locations (unreliable internet) needs an ML solution to detect package dimensions in real-time video footage. Model training is done in SageMaker AI. Goal: real-time decision-making without relying on constant cloud connectivity. ? Question: Which of the following solutions would best meet the company ?s needs?. 

Deploy a Convolutional Neural Network (CNN) in SageMaker AI using Amazon Kinesis Video Streams to analyze the video footage in real time. Use Amazon EventBridge to trigger downstream actions for routing packages based on the detected dimensions. 
Train the model using SageMaker AI and deploy it to Amazon Elastic Kubernetes Service (Amazon EKS) clusters running in each factory. Use Amazon SQS to queue routing decisions and send them to the cloud for processing.
Use Rekognition Custom Labels to train the model and deploy it using Amazon EC2 instances at each factory. Use Amazon EventBridge to monitor inference results and trigger routing actions
Use SageMaker?s built-in Object Detection algorithm to train the model. Deploy the trained model to an AWS IoT Greengrass core with AWS Lambda handling the decision logic at the factory

Show Answer

Question # 9

 Question on AI Agent/Workflow Orchestration ? Category: AIP – Implementation and Integration. ? Scenario: A customer service assistant needs to handle complex order inquiries, maintain conversation context across sessions, and securely update order records (execute actions). ? Question: Which solution best satisfies the company ?s requirements?

Use Amazon Lex V2 to build a conversational chatbot for customer interactions and store conversation transcripts in Amazon S3 for historical analysis.
Use Amazon Titan Text G1 for conversation handling and maintain customer session states in a Python dictionary within the application memory for short-term interactions. 
Use Amazon Kendra to search for answers to product manuals and FAQs. Combined with AWS Lambda to manage refund requests and data updates.
Use Amazon Bedrock AgentCore to develop an AI agent capable of reasoning, planning, and executing workflows for order management. Integrate the agent with Amazon DynamoDB to store and retrieve customer session data, order history, and interaction context for each user conversation. 

Show Answer

Question # 10

Question on BERT Fine-Tuning/Transfer Learning ? Category: AIP – Foundation Model Integration, Data Management, and Compliance. ? Scenario: An email filtering system needs to fine-tune a pre-trained BERT model for spam detection using a labeled email dataset (binary classification). Goal: correctly load the pretrained weights and use them as the initialization point for fine-tuning without full retraining. ? Question: Which approach will correctly initialize the BERT model to achieve this requirement?.

Load the pretrained model weights for every layer and place an external classifier on top of the primary model output vector. Train the newly added classifier with the labeled dataset.
Use the pretrained model weights for all transformer layers and attach a second classifier layer in parallel with the existing output layer. Train only this additional classifier using the labeled dataset. 
Initialize the model with pretrained weights, convert the output layer into a multi-task classifier that predicts multiple text classes beyond spam detection, and train this classifier using the labeled dataset.
 Apply pretrained model parameters across all layers, then discard the existing final layer. Introduce a custom classifier and train it using the labeled data for spam detection.

Show Answer

Question # 11

 Question on Churn Prediction (Automation/Explainability) ? Category: AIP – Implementation and Integration. ? Scenario: A retail team needs an automated way (minimal manual effort) to build a model to predict customer churn and identify the most relevant features contributing to the prediction (explainability). Question: Which of the following solutions will best fulfill these requirements while minimizing manual effort?.

Use SageMaker Data Wrangler to automatically train a churn prediction model and rely on its quick model visualization feature to generate accurate importance scores for deployment decisions.
Use SageMaker Ground Truth to label customer churn data, then build a custom TensorFlow model to predict churn and analyze feature weights posttraining
Use the k-means algorithm in SageMaker AI to cluster customers based on purchasing patterns. After clustering, use the resulting clusters to predict churn based on customer behavior
 Leverage SageMaker Autopilot to automatically train a classification model for forecasting customer churn. Then, utilize insights from SageMaker Clarify to determine which features most significantly influence the predictions

Show Answer

Question # 12

Question on Endpoint Scaling Policy ? Category: AIP – Operational Efficiency and Optimization for Generative AI Applications. ? Scenario: A recommendation endpoint experiences significant delays during predictable high-traffic sales events, resulting in poor user experience. The goal is to adjust the target tracking scaling policy to proactively ensure sufficient capacity and prevent latency issues during these peak periods. ? Question: Which solution will best meet the requirements?.

Increase the instance size of the SageMaker endpoint to a larger instance type to accommodate higher traffic during sales events.
Implement a step scaling policy for the SageMaker inference endpoint that scales based on resource utilization metrics such as CPU and memory usage.
Use AWS Lambda to periodically restart the SageMaker endpoint during peak traffic to refresh instance performance.
Configure a scheduled scaling policy to increase the capacity of the SageMaker inference endpoint before the sales events begin. 

Show Answer

Question # 13

 Question on API Token Rotation ? Category: AIP – AI Safety, Security, and Governance. ? Scenario: A fraud detection system relies on external APIs, and the security policy requires rotating API tokens every 3 months. Goal: automate token rotation, ensure secure token storage, and maintain continuous operation without downtime. ? Question: Which solution will best address these requirements?.

Use AWS Key Management Service (AWS KMS) with customer-managed keys to store the tokens and rely on Amazon EventBridge to trigger rotation events. 
Use AWS Systems Manager Parameter Store to store the tokens and rely on an AWS Lambda function for automatic rotation. 
Use AWS Secrets Manager to store the tokens, monitor API usage with AWS CloudTrail, and rely on Amazon EventBridge to trigger token rotation. 
Use AWS Secrets Manager to store the tokens and rely on an AWS Lambda function to perform the rotation process. 

Show Answer

Question # 14

Question on Safe Deployment Strategy ? Category: AIP – AI Safety, Security, and Governance. ? Scenario: A new model version for credit default risk prediction needs to be deployed to a SageMaker real-time inference endpoint. Previous deployments experienced latency spikes and failures. Goal: minimize downtime and mitigate performance degradation risk by using a deployment strategy that offers safe rollout and automatic rollback capabilities. ? Question: Which deployment configuration best meets these requirements?.

 Use SageMaker batch transform to validate the new model offline. Promote directly to full production using a single update event. 
Use a shadow testing deployment to send duplicate inference requests to the new model. Log results for later comparison, without affecting live predictions. 
Deploy both models using a multi-model endpoint configuration. Dynamically select the model version at runtime based on an API request parameter.
Configure a blue/green deployment with canary traffic shifting and a traffic size of 10%. Gradually route requests to the new model while maintaining the existing version as a fallback

Show Answer

Question # 15

 Question on Secure S3 Access (IAM Role) ? Category: AIP – AI Safety, Security, and Governance. ? Scenario: A SageMaker notebook instance needs appropriate permissions to read training data from one S3 bucket and write model artifacts, logs, and evaluation results to a different S3 bucket. Goal: grant secure and proper access control. ? Question: Which approach should be used to securely enable this access?

Define a bucket policy on the S3 bucket that allows the SageMaker AI notebook instance by its ARN to perform s3:GetObject, s3:PutObject, and s3:ListBucket actions
Use AWS IAM identity federation to provide temporary access to the S3 bucket by configuring the SageMaker notebook instance to assume a federated role for accessing the data.
Create an S3 access point for the SageMaker notebook instance, granting it access to the necessary data, and configure the access point to allow only the required actions (s3:GetObject, s3:PutObject, and s3:ListBucket).
Allow the SageMaker notebook instance to perform s3:GetObject, s3:PutObject, and s3:ListBucket operations by attaching a policy to its associated IAM role that grants access to the designated S3 buckets

Show Answer

Question # 16

 Question on Secure On-Premises to S3 Transfer ? Category: AIP – AI Safety, Security, and Governance. ? Scenario: A healthcare organization must upload non-sensitive data from an onpremises Microsoft SQL Server database to Amazon S3 for model retraining. Sensitive patient records must never leave the data center. All transfers must be done securely over an IPsec connection. ? Question: Which solution will satisfy the given requirements?.

Configure AWS Database Migration Service (AWS DMS) to replicate nonsensitive data from the Microsoft SQL Server database into S3, and transfer it over an IPsec connection.
Utilize Amazon Data Firehose to stream non-sensitive transaction data into S3. Ensure that the data transfer happens over an IPsec-protected connection, and leverage AWS Lambda to filter out sensitive data before uploading
Set up an AWS Glue job to connect to the Microsoft SQL Server database, extract only the non-sensitive data, and transfer it to S3 over an AWS Site-to-Site VPN connection for model retraining
Use AWS Transfer Family to securely transfer the entire database backup file to S3, and then use AWS Glue to filter sensitive data from the S3 bucket. 

Show Answer

Question # 17

Question on A/B Testing (Multi-Variant Endpoints) ? Category: AIP – Testing, Validation, and Troubleshooting. ? Scenario: Multiple recommendation models must be evaluated using A/B testing in production. The system must route live inference traffic, monitor real-time engagement metrics, and seamlessly direct 100% of traffic to the best-performing model with minimal operational overhead. Question: Which solution will meet these requirements in the most operationally efficient way?.

Deploy the models on Amazon EC2 instances behind an Application Load Balancer (ALB) to perform A/B testing, then manually adjust the ALB weights when a model shows higher engagement. 
 Create a separate Amazon SageMaker AI endpoint for each model and configure Amazon API Gateway to distribute traffic for A/B testing based on weighted routing rules
Use AWS CodeDeploy with blue/green deployment strategies and an Application Load Balancer (ALB) to alternate traffic between model versions during A/B testing. Gradually route 100% of traffic to the model with the highest engagement metrics.
Use Amazon SageMaker AI multi-variant endpoints to deploy all model versions behind a single endpoint. Configure traffic weights for A/B testing and update routing to send all inference requests to the best- performing model once identified.

Show Answer

Question # 18

Question on Bias Detection/Explainability (Recommendations) ? Category: AIP – AI Safety, Security, and Governance. ? Scenario: A recommendation model shows specific product categories disproportionately to customers from certain regions. The organization must verify if the imbalance is due to dataset bias or model prediction bias , and must generate automated reports explaining feature influence for compliance and transparency. ? Question: Which solution will meet these requirements?.

Use Amazon Personalize to automatically adjust recommendation weights in real-time to reduce category bias without performing explicit fairness evaluation.
Use SageMaker Data Wrangler to manually rebalance the dataset by filtering and transforming data before retraining the recommendation model.
Use SageMaker Model Monitor to track endpoint metrics such as latency, drift, and accuracy without analyzing model bias or feature influence
Use SageMaker Clarify to detect and measure data bias, evaluate model fairness, and generate feature attribution explainability reports for compliance and transparency.

Show Answer

Question # 19

Question on Secure API Traffic (VPC Endpoint) (Select TWO) ? Category: AIP – AI Safety, Security, and Governance. ? Scenario: An ML pipeline uses a SageMaker Service API VPC interface endpoint in a public subnet. The team must ensure that only specific Amazon EC2 instances and IAM users can invoke SageMaker API operations through that endpoint. ? Question: Which combination of actions should the team take to secure the traffic to the SageMaker Service API? (Select TWO.).

Enable private DNS for the VPC endpoint to ensure that traffic remains within the VPC.
Enable VPC Flow Logs to monitor traffic patterns. Use AWS Lambda to automatically block unauthorized access to the SageMaker API endpoint. 
Deploy an additional VPC endpoint for SageMaker AI Runtime to isolate inference traffic
Attach a custom VPC endpoint policy that explicitly grants access to selected IAM identities
Configure the security group linked to the endpoint network interface to allow traffic only from approved instances. 

Show Answer

Question # 20

Question on Multimodal RAG (BDA) ? Category: AIP – Implementation and Integration. ? Scenario: A publishing company needs a generative AI chatbot to answer queries based on a repository of documents, images, audio, and video content. Goal: Process all media types, index them for semantic search, and use the retrieved context to ground a foundation model (LLM) hosted on SageMaker AI. ? Question: Which approach best meets these requirements?.

 Employ Bedrock Data Automation (BDA) to process all media types and directly supply the structured output into a foundation model via SageMaker AI, bypassing the use of a knowledge base
Use Bedrock Data Automation (BDA) to process the media files, store the raw content in Amazon S3, and deploy a custom AWS Lambda function to create your own vector database outside of Bedrock Knowledge Bases before passing it to SageMaker AI.
Utilize Bedrock Data Automation (BDA) to process media files, analyze the structured content using Amazon Comprehend for entity and sentiment extraction, and then forward the results to a foundation model hosted on Amazon EC2 for generating answers
Leverage Bedrock Data Automation (BDA) to process documents, images, audio, and video, index the results in Bedrock Knowledge Bases for semantic search, and then input the retrieved context into a foundation model via SageMaker AI for response generation

Show Answer

Question # 21

 Question on Feature Attribution Drift Monitoring ? Category: AIP – Testing, Validation, and Troubleshooting. ? Scenario: A claims automation system uses SageMaker AI, predicting claim approval based on vehicle damage severity and other features (age, mileage). The model must be continuously monitored for feature attribution drift in production (i.e., if the model starts prioritizing less relevant features like vehicle age over damage severity). ? Question: Which solution should be implemented?.

Deploy SageMaker Clarify to perform bias and explainability analysis on the training dataset. Use Amazon CloudWatch to alert if Clarify reports significant changes in feature attribution or fairness metrics. 
Implement a baseline for model quality using the ModelQualityMonitor class. The baseline will evaluate key performance metrics such as accuracy and recall, with periodic checks to identify any significant shifts in model performance. Set up an Amazon CloudWatch if the model ?s quality metrics diverge from the baseline
Enable SageMaker DataCapture to log inference inputs and outputs. Build a custom pipeline to analyze feature distributions and model responses over time. Use Amazon CloudWatch to alert when significant shifts in input patterns or predictions are detected.
 Use ModelExplainabilityMonitor class with a SHAP-based baseline to detect feature attribution drift in production. Regularly compare how the model assigns importance to input features against the baseline, and configure Amazon CloudWatch to alert stakeholders when attribution values drift beyond acceptable thresholds

Show Answer

Question # 22

. Question on Real-Time Fraud Detection (Minimal Overhead) ? Category: AIP – Implementation and Integration. ? Scenario: A digital payments provider is experiencing fraudulent transactions, especially from new accounts performing high-value payments. The existing batch model in SageMaker cannot flag activity quickly enough. Goal: real-time fraud detection that can automatically assess and reject fraudulent transactions at the moment of occurrence, requiring minimal operational effort. ? Question: Which option satisfies these requirements?. 

Use Amazon Lookout for Vision to detect anomalies in uploaded transaction receipt images and classify them as fraudulent or legitimate
Use Comprehend to extract entities from transaction metadata and forward them to SageMaker AI to retrain a fraud detection model.
Use SageMaker AI to train a new supervised model for fraud detection and deploy it on Amazon EC2 using custom inference code.
. Use the Amazon Fraud Detector prediction API to automatically approve or deny transactions that are identified as fraudulent. 

Show Answer

Question # 23

Question on Semantic Embeddings/RAG ? Category: AIP – Foundation Model Integration, Data Management, and Compliance. ? Scenario: A research team needs a mechanism to represent user queries and internal documents as semantic embeddings to capture contextual relationships. The solution must be fully managed, scalable, and integrate easily with Bedrock AI agents for downstream RAG workflows. ? Question: Which approach best satisfies these requirements?.

Implement Amazon Kendra to index research documents, support naturallanguage queries, and let AI agents retrieve relevant results using the managed semantic-search and ranking capabilities of the service.
 Configure SageMaker Data Wrangler to preprocess textual data, extract engineered features through clustering, and allow AI agents to analyze document similarity within structured datasets and grouped content.
Deploy SageMaker JumpStart to fine-tune and host a pre-trained language model for summarization and text generation, integrating it with AI agents for enhanced content discovery workflows.
. Leverage Amazon Titan Text Embeddings in Bedrock to convert text into semantic vectors and store them in Amazon OpenSearch Service for context-aware retrieval and reasoning by AI agents

Show Answer

Question # 24

Question on Text Preprocessing for Word2Vec (Select THREE) ? Category: AIP – Foundation Model Integration, Data Management, and Compliance. ? Scenario: Training a Word2Vec-style model on SageMaker AI requires preparing a dataset of over one million sentences with inconsistent casing, mixed encodings, and minor typographical errors (e.g., “An Apple a DAY Keeps the doctor Away”). Goal: ensure consistent sanitization, reproducibility, and embedding quality. ? Question: Which of the following operations should be implemented in the preprocessing phase to correctly sanitize and prepare the dataset for embedding and downstream predictions? (Select THREE.).

. Apply part-of-speech tagging to identify grammatical elements and retain only the verbs and nouns. 
Replace every word with its corresponding synonym using a lexical database before tokenization
Normalize the text by converting every word in the sentence to lowercase.
Exclude common non-informative words from the dataset using an English stop-word dictionary
Segment the sentence into individual word units through tokenization. 
Convert all tokens into fixed-length character n-grams before Word2Vec training to capture subword features.

Show Answer

Question # 25

Question on Asynchronous Inference (Long Processing Time) ? Category: AIP – AI Safety, Security, and Governance. ? Scenario: A fraud detection model must run within a private VPC (no public internet). The model handles payloads of 6 MB to 11 MB and requires a long inference execution time (18–22 minutes) per request. Cost minimization is also a requirement while supporting a request/response style interaction. Question: Which solution is the most suitable approach to satisfy this requirement?.

Configure a SageMaker Batch Transform job to run within private subnets and attach the appropriate VPC configuration parameters during endpoint creation
Use SageMaker multi-model endpoint architecture inside private subnets with VPC configuration applied as part of endpoint deployment procedures. 
Use SageMaker Neo compiled model packaging and deploy the compiled artifact to a SageMaker real-time inference endpoint inside private subnets using VPC configuration.
Deploy a SageMaker asynchronous endpoint inside private subnets and include VPC configuration parameters during endpoint creation. 

Show Answer