From Bubble exports to a platform the company runs on.
Joined as an analytics engineer. Seven months later the company runs on the platform I built, and a new product went from idea to production in four weeks.
Where it started.
SharpStakes runs hundreds of ambassador campaigns with sportsbook and casino partners. When I joined, the data lived inside Bubble: reports came from CSV exports, proof approvals took days, and nothing tied payouts, meetings and partner revenue together. Leadership had no single view of the money.
How I worked it.
A real data layer
Stood up Supabase as the source of truth, synced from Bubble in real time with a nightly backfill, and connected Dots, QuickBooks, Mercury, Calendly, Twilio and scrapers for partner portals that have no API.
An internal platform
Built a 13-module internal app for sales, fraud, finance and leadership: rep workstations, reporting and margin, fraud review, partner caps, meetings and payouts. It took over most of what the Bubble team did; one developer maintains Bubble through the migration, and I manage them.
Controls that run themselves
Screenshot fraud checks on every proof, automatic brand pauses when a partner cap fills, and a Slack bot that nudges reps about unconfirmed deposits. One rep recovered $3K+ in the bot's first week.
A new product line
Owned a new real-money consumer product from spec to production in four weeks. I wrote the spec, made the product, compliance and payments calls, and built it with AI coding agents under a test-first process: 2,500+ automated tests, every change reviewed.
Building the team
Co-running hiring for full-stack engineers: wrote the role, built a take-home test with an automated grader, and interview candidates for the team I'll manage.
What shipped.
- Median proof approval went from 3 days to 12 minutes across 10,000+ approvals.
- Lost-commission rate is down 57%.
- 1,374 automatic brand pauses across 174 brands since July, with no one watching a dashboard.
- Leadership tracks C$4.1M in partner revenue, margin and cash flow in one place.
Want to compare notes?
Always happy to talk about data platforms, internal tools and building with AI agents.