Background
The first problem I shipped in production. Chinese students applying to universities abroad needed help with Personal Statements, Research Proposals, and program fit. The traditional path — paid agencies, single-fee packages — was heavy and expensive. I wanted a lighter model.
What I built
- **The light-asset model.** Overseas professors, lecturers, and senior students signed agreements to mentor on a per-session basis. No full-time hires, no long-form contracts.
- **The mentor library.** A structured catalog covering region preferences (US, UK, EU, Asia), major preferences, advisor seniority, and rate. Searchable from both sides.
- **The matching system.** Domestic students browse, request, and meet their chosen mentor 1-on-1. Built for time-zone friction, payment friction, and expectation alignment.
- **The admin surface.** A basic back-office to manage the mentor roster, contracts, payouts, and disputes.
- **The open page.** A public discovery surface for new students to find their match without going through a sales call.
What I learned
This is where I built my first production system, solo. The hardest part was not code — it was the cross-border coordination: time zones, language, payment, expectations. I had to design the experience so that an overseas professor and a Chinese student could meet inside 48 hours without either side being the bottleneck. That instinct — design around the bottleneck, not around the feature — has shaped every system I have built since.
Outcome
The platform ran through the end of my tenure. Many of the early students went on to receive offers from target schools; many of the early mentors stayed on as returning advisors. And the pattern — independent contributors, coordinated by a small system — is the same pattern I later used to build the GUILD Digital Operations Ecosystem.