All work

Analytics SaaS

Project Atlas

A real-time analytics dashboard for B2B customers, processing 40 million events a day without slowing down.

Built with
React, Node.js, Redis, WebSockets
Outcome
Reports that took 8 seconds now load in under 100ms.
Project Atlas artwork
40Mevents a day
<100msqueries
54%lower cloud costs

The problem

A B2B analytics platform shared by many client companies kept timing out during working hours. Its database ran at 100% CPU on unindexed time-based reports, and a 30-day dashboard took more than 8 seconds to load.

What I did

I built an ingestion pipeline with Node.js, Redis Streams, and PostgreSQL materialized views, so heavy reports are prepared ahead of time. Dashboards get live updates over WebSockets instead of asking the server again and again. Tuned connection pools, automatic query timeouts, and circuit breakers keep one slow query from taking the rest down.

The result

Queries went from 8,200ms to under 100ms across more than 40 million events a day. Server costs fell 54% because the database stopped triggering extra machines, and churn tied to slowness fell to zero.