From Raw Footage to Publish-Ready: Building a Serverless AI Video Pipeline on AWS

Creating high-quality video content at scale is time-consuming, resource-intensive, and heavily reliant on manual editing workflows. For modern newsrooms and content creators, the bottleneck isn’t capturing footage - it’s everything that comes after.

In this session, I’ll walk through the architecture of a fully automated, serverless video pipeline built on AWS that transforms raw video into publish-ready content. The pipeline combines AI-powered transcription enhancement, detailed per-frame visual analysis, automated content generation, and intelligent editing decisions - all orchestrated without a single server to manage.

We’ll explore how AI Agents coordinate the end-to-end workflow with Step Functions, how Rekognition and Bedrock Data Automation handle deep video and frame-level understanding, how Bedrock foundation models drive content generation and editing logic, and how MediaConvert handles the heavy lifting of video processing. S3 ties it all together as the central media store.

Whether you’re a developer curious about event-driven media architectures or an architect evaluating AI services for content automation, this session offers real-world lessons from building a production-grade pipeline that makes video creation faster, smarter, and scalable.