Operations platform, web
The operations platform behind a studio of iOS apps: revenue and subscriptions from RevenueCat and App Store Connect, ad spend, creator payouts and alerts, in one Postgres schema of 63 tables.

- Database
- 63 tables
- Migrations
- 88
- Code
- ~95k lines
The problem
Studios shipping a portfolio of apps lose hours every week stitching together RevenueCat, App Store Connect, TikTok analytics, and ad spend across separate dashboards. By the time a paywall regression shows up in a monthly review, the damage is done.
What I built
Viral OS is the operations dashboard for a B2C app studio. It ingests App Store Connect (sales + metadata), RevenueCat (101 webhook event types), TikTok (via a Playwright scraper running on GitHub Actions, not the paid API), and manual ad spend; computes metrics that aren't in any individual tool: predicted LTV (geometric with realized-churn fallback), LTV/CAC, cohort retention, app-health score (0 to 100), lifecycle stage; and fires Slack/Discord/email alerts the same day a paywall regression or refund spike appears.
One Supabase-backed dashboard pulls every data source into a unified schema. Hand-rolled detectors (EWMA + 2σ on revenue + refunds + MRR drops + rating drops + rejected builds) catch anomalies same-day. Gemini 3 Pro is wired into every surface via tool-use, not stuffed context. The AI queries the live DB instead of getting fed it.

