Answer: Use Amazon SageMaker Serverless Inference to deploy the model.

## Detailed Explanation ### Requirements Analysis The question specifies three key requirements: 1. **Host the ML model** - Provide infrastructure to store and run the model 2. **Serve predictions** - Enable real-time inference capabilities 3. **No management of underlying infrastructure** - Fully managed, serverless solution ### Option Analysis **A: Amazon SageMaker Serverless Inference** - This is the optimal solution because: - **Purpose-built for ML hosting**: SageMaker is AWS's dedicated service for machine learning lifecycle management - **Serverless architecture**: Automatically provisions, scales, and manages compute resources without user intervention - **Real-time inference**: Designed specifically for serving predictions with low latency - **Infrastructure abstraction**: No need to manage EC2 instances, scaling policies, or load balancers - **Integration ready**: Provides HTTPS endpoints that web applications can easily consume **B: Amazon CloudFront** - Not suitable because: - CloudFront is a Content Delivery Network (CDN) for caching and delivering static/dynamic content - Not designed for hosting or executing ML models - Cannot perform inference computations **C: Amazon API Gateway** - Not suitable because: - API Gateway is an API management service for creating, publishing, and monitoring APIs - While it can route requests to backend services, it doesn't host ML models itself - Would require additional infrastructure (like Lambda or EC2) to actually run the model **D: AWS Batch** - Not suitable because: - AWS Batch is designed for batch processing and job scheduling - Not optimized for real-time inference serving - Requires infrastructure management and doesn't provide direct web-accessible endpoints ### Best Practices Alignment For ML model deployment with infrastructure abstraction, AWS best practices recommend: 1. **SageMaker Serverless Inference** for unpredictable or intermittent workloads 2. **SageMaker Real-time Inference** for high-throughput, consistent workloads (requires some infrastructure management) 3. **SageMaker Multi-model Endpoints** for hosting multiple models efficiently Given the requirement for zero infrastructure management, SageMaker Serverless Inference is the only AWS service that meets all criteria without requiring any provisioning, scaling, or maintenance of compute resources.

Author: LeetQuiz Editorial Team