Context. Capcod's DocGen microservice turns HTML templates into PDFs. The rendering step is bursty and CPU-heavy, so I moved it out of the always-on service and onto a dedicated AWS Lambda.

• A container-image Lambda on .NET 8 (custom runtime via LambdaBootstrapBuilder), bundling a Lambda-optimized headless Chromium that does the HTML-to-PDF rendering.
• The CI/CD as code: a CloudFormation template that defines a CodePipeline and CodeBuild to build the Docker image, push it to ECR, pull private dependencies from CodeArtifact, and update the function.
• The client side: the main .NET service invokes the Lambda through the AWS SDK (IAmazonLambda). I handled Lambda's 6 MB synchronous payload limit by base64-encoding the PDF in the response.

This is real production AWS work I owned end to end: Lambda, ECR, CodePipeline, CodeBuild and CodeArtifact, all in eu-central-1. It's the piece I point to first for serverless on AWS.

AWS Lambda (container image) · .NET 8 / C# · headless Chromium · Docker / ECR · CodePipeline / CodeBuild · CloudFormation · CodeArtifact.