BridgeSync: an AI progress-tracking SaaS built end-to-end, billing included
My university final-year project, built solo in four months: a SaaS that turns raw GitHub activity into client-ready progress reports (hybrid rule + LLM matching of pull requests to requirements), with auth, plans, and Stripe subscription billing done properly.
- Role
- Solo full-stack: architecture, frontend, backend, and payments
- Stack
- Next.js 16 · Convex · Clerk · Stripe · Vercel AI SDK
- 1,809Automated tests passing
- 46 / 63App routes / data tables
- 93.3%PR-to-requirement rule-match accuracy (hand-labelled)
The problem
Clients constantly ask how their project is going; developers hate writing status updates. Turning raw GitHub activity into client-readable progress (without leaking internal mess) is a real product problem, stacked on top of the usual SaaS hard parts: auth, plans, and recurring billing.
The approach
- Extracts requirements from uploaded SRS documents, then matches incoming GitHub pull requests to them with a hybrid rule + LLM pipeline, confidence-graded, with human review on low-confidence matches (rule layer: 93.3% accuracy on a hand-labelled set).
- Tracks developer progress passively from GitHub App webhooks (zero manual status updates), and an approval-gated boundary keeps internal state separate from what clients ever see.
- Built the entire platform solo: Next.js 16 with a Convex real-time backend, Clerk auth, Stripe subscription billing, and a client-facing PWA. 46 routes over a 63-table schema, in four months.
The results
A pre-launch SaaS with production mechanics proven the engineering way (1,809 automated tests passing and a real billing integration), proof of owning an entire product, including the parts side-projects skip. User metrics will be published once it launches, not before.
Have a process like this eating your team's time?
Book a 15-minute fit call →