Back to work
AI Systems · ProjectLive · Public
LevelMyAudio
Professional audio cleanup, democratized — a 10× workflow reduction.
Summary
An AI-powered audio processing pipeline that brings professional leveling to any creator, collapsing a traditionally manual workflow by roughly 10×. Built solo.
The Problem
Professional audio leveling is slow, expensive, and gatekept by expertise. Creators need broadcast-quality output without an engineer in the loop.
Approach
- Built an upload-to-leveled pipeline targeting the −16 LUFS broadcast standard.
- Ran heavy processing on elastic Fargate workers driven by an SQS queue.
- Notified users asynchronously so the UI never blocks on long jobs.
- Defined the whole environment in Terraform.
Architecture
- 01FastAPI backend with a Next.js frontend on Vercel.
- 02AWS ECS Fargate workers for FFmpeg-based processing.
- 03SQS job queue decoupling upload from processing.
- 04Terraform IaC for the full environment.
Key Tradeoffs
Fargate over always-on workers
Audio jobs are bursty; serverless containers meant no paying for idle CPU between uploads.
Async job + email over synchronous processing
Long encodes can’t block a request; a queue plus notification keeps the UX responsive and the system resilient.
Stack
FastAPINext.jsAWS ECS FargateSQSFFmpegTerraform
Outcome
A roughly 10× reduction in the traditional leveling workflow, at the −16 LUFS broadcast target.