Atlas — a GTM data platform from zero to scale
A unified system that lets revenue teams enrich, score, and route thousands of accounts in real time — without writing a single SQL query.
The problem
Revenue teams at Moonshot were drowning in disconnected spreadsheets. Account data lived in five tools, enrichment was a manual copy-paste ritual, and by the time a lead was scored it was already cold. Leadership wanted one place where data flowed in, got enriched, and triggered action automatically.
Architecture & approach
I designed Atlas around an event-driven core: a Kafka backbone ingests records, a fleet of Go workers fan out enrichment calls to third-party APIs, and results land in Postgres with a Redis cache layer for hot reads. The frontend is a React app that treats the whole thing like a live spreadsheet — every cell is reactive, so the moment data resolves, the UI updates without a refresh.
The hardest part
Making 12 million monthly enrichments feel instant. I built a speculative-execution layer that pre-warms enrichment for accounts a user is likely to open next, and a backpressure system so a slow third-party API never stalls the pipeline. We got p95 query latency down to 340ms even under peak load.
Outcome
Atlas went from an internal experiment to the daily driver for the entire GTM org — adoption grew from 6 to 48 people in two quarters. Manual data work dropped by an estimated 70%, and the platform became the foundation the company now sells to its own clients.