Atlas Analytics
Self-serve product analytics with sub-second queries over billion-row events
A multi-tenant analytics platform where teams define events, build funnels, and query aggregated results without writing SQL. The interesting work was in the query layer: a columnar store, incremental rollups, and a caching strategy that kept p95 under 400ms while writes stayed continuous.
- Incremental hourly rollups cut p95 query latency from 4.2s to 380ms
- Virtualised data grid renders 100k rows without measurable frame drops
