The correct answer is: Use a Lambda@Edge function that includes an image processing package, and integrate it with CloudFront behaviors handling image requests.

Lambda@Edge allows developers to run functions at AWS edge locations in response to CloudFront events. By attaching an image manipulation library such as Sharp to the function, images can be dynamically resized, compressed, or reformatted based on the request parameters, including the User-Agent. This solution offloads all image processing to the edge, reduces latency, and removes the need for a dedicated image server. It also scales automatically and has minimal operational overhead once deployed.

The suggestion to set up a CloudFront origin request policy to modify image responses based on User-Agent headers and generate resized assets is incorrect because request policies control which headers, cookies, or query strings are forwarded to the originthey dont perform transformations or resizing. No automatic image manipulation is performed by CloudFront just by using request policies.

The choice to deploy a third-party image processing tool on a dedicated EC2 instance introduces significant management burden. You must manage the server, monitor performance, handle scaling, security, and patching. This contradicts the requirement to minimize operational overhead, and this design doesn't benefit from CloudFronts edge processing power.

The option to apply a CloudFront response headers policy to resize images and automatically serve different formats based on the User-Agent header is not valid. Response header policies are used to control headers returned to clients, such as setting cache-control or CORS headersthey do not modify the content of the response or perform image transformation.

This question is important because it evaluates your ability to use advanced AWS features like Lambda@Edge for real-time content transformation, which is a frequent pattern for optimizing content delivery and personalization. The challenge here is understanding the role and limits of various CloudFront policies and recognizing when edge computing provides a strategic advantage.

To answer these types of questions, focus on keywords like dynamically, based on client, and low operational overhead. These indicate the need for serverless, auto-scaling solutions at the edge, and rule out traditional server-based approaches. Lambda@Edge often comes up as the optimal tool in such cases, especially for real-time personalization or transformation.

Full AWS Practitioner Certification Question