Cloud spend cut 50%, 24,000 errors a day eliminated
The problem. A carbon accounting platform was outgrowing its infrastructure: a rising cloud bill, thousands of daily database errors, and data pipelines held together by an external scheduler.
What we built. We led backend engineering across an integration-heavy platform: accounting system and energy data integrations, real-time ingestion on distributed task queues, and scheduled pipelines migrated to serverless functions.
- Cloud costs halved through database right-sizing, query optimization, and 36 missing indexes
- Around 24,000 daily SQL errors traced to a single lookup bug and eliminated
- Database reads cut from 67M to under 20M per month
- P99 database CPU brought back to baseline