Sample Questions from AWS Certified AI Practitioner (AIF-C01)

Preview a few questions below — answers are revealed when you take the exam.

1. A team needs to deploy a machine learning model on AWS that requires low latency and high throughput. What approach should they take to optimize performance?

   • Deploy the model on Amazon SageMaker using a single ml.m5.large instance and configure auto-scaling based on CPU utilization.
   • Deploy the model on Amazon EC2 using multiple ml.p3.2xlarge instances with Amazon Elastic Fabric Adapter (EFA) for high-performance computing.
   • Deploy the model on AWS Lambda with provisioned concurrency to handle varying traffic patterns efficiently.
   • Deploy the model on Amazon SageMaker using a multi-model endpoint to serve multiple models simultaneously with optimal resource utilization.

2. Consider the situation where a company wants to implement a recommendation system using AWS services. Which method solves it best?

   • Utilize Amazon Personalize to create and deploy a recommendation system without requiring machine learning expertise.
   • Build a custom recommendation system using Amazon SageMaker and train it with collaborative filtering algorithms.
   • Use Amazon Pinpoint to segment customers and send personalized recommendations via email campaigns.
   • Implement a recommendation system using Amazon Redshift to analyze customer purchase history and generate recommendations.

3. Which option best evaluates the effectiveness of a natural language processing (NLP) model deployed on AWS for sentiment analysis?

   • Conduct A/B testing with different model versions and measure the accuracy, precision, and recall of sentiment predictions.
   • Analyze the model's performance metrics such as F1 score, confusion matrix, and ROC curve to assess its effectiveness.
   • Gather user feedback through surveys and interviews to evaluate the model's impact on customer satisfaction and engagement.
   • Monitor the model's latency, throughput, and resource utilization to ensure optimal performance and cost-efficiency.

4. How should a professional assess the security posture of an AWS-based AI/ML pipeline?

   • Perform a security audit using AWS Well-Architected Framework and identify any vulnerabilities or compliance gaps.
   • Conduct penetration testing and vulnerability assessments to identify potential security risks and weaknesses.
   • Review the access controls, encryption mechanisms, and data handling practices implemented in the pipeline.
   • Implement AWS Identity and Access Management (IAM) policies to enforce least privilege access and restrict unauthorized actions.

5. What strategy should be applied when optimizing the cost of running machine learning workloads on AWS?

   • Utilize AWS Cost Explorer to analyze spending patterns, identify cost optimization opportunities, and implement cost-saving measures.
   • Right-size EC2 instances based on workload requirements, leverage spot instances for cost-effective compute, and optimize data storage costs.
   • Implement AWS Savings Plans and Reserved Instances to commit to long-term usage and benefit from discounted pricing.
   • Automate resource provisioning and scaling using AWS CloudFormation and AWS Auto Scaling to optimize resource utilization and reduce costs.