Next-Gen PDF Rendering: Scaling to 1 Million Documents in 10 Minutes with Serverless Rust

Generating documents at scale is a common but deceptively complex challenge, especially in regulated industries like finance where millions of reports are required within tight time windows. A naive approach of throwing thousands of vCPUs at the problem is not only prohibitively expensive but often fails to scale gracefully. This session presents a real-world case study on how we architected a PDF rendering pipeline to generate 1 million documents in under 10 minutes for under a dollar, a 95% cost reduction compared to conventional cloud-native solutions.

We'll dive deep into the architecture of a fully asynchronous, event-driven pipeline in the cloud, all managed with Terraform. The core of our solution ditches slow, memory-hungry engines like Puppeteer or LaTeX in favor of Typst, a modern, high-performance typesetting system. By harnessing its speed within a native Rust Lambda function, we eliminated runtime overhead and achieved hot-invocation processing times of < 40ms per document.