DESYGNER
100M+ records, 1.2M deletions a day, 20% off the AWS bill
Scaled an asset-rendering platform to 100M+ records, lifted render performance 40%, and cut EC2 spend 20%.
Problem
An asset-management and rendering platform served millions of users against 100M+ digital records. Rendering was a bottleneck, GDPR-style deletions at scale were unsolved, and EC2 spend was climbing without a ceiling.
Approach
I redesigned the queue systems and database queries behind the Node.js/TypeScript/SharpJS rendering path, then built a multithreaded PHP + Redis pipeline to delete users and all their associated media safely at volume. On the infrastructure side I rightsized instances, introduced autoscaling and workload scheduling, and containerised the whole platform with Docker, Terraform, and CI/CD, backed by Datadog, Sentry, and Grafana.
Result
Render performance rose 40%, the deletion pipeline cleared 1.2M users/day including their images and videos, and EC2 spend fell roughly 20%, a meaningful cut to the monthly cloud bill. Onboarding ramp-up for new engineers dropped 50% through structured mentorship and documentation.
Scale problems are rarely one big fix, they’re a chain of bottlenecks. The interesting engineering here was the deletion pipeline: doing destructive work, fast, across millions of records and their media, without dropping the wrong thing or falling over under load. Multithreaded PHP coordinated through Redis turned a compliance liability into a routine background job.