LinKross

Period: 2026
Team project (5 members) · SeSAC “AI-Driven Data Project Course for Service Planners/PMs”
My role: UI/UX and freelancer-experience design
Code: github.com/kimminjunnn/LinKross
Problem
Generative AI makes it fast to start building, but a prototype is not a working MVP. Early-stage founders often cannot afford an in-house development team, and overseas outsourcing is cheaper but risky at every stage: choosing a developer, agreeing on requirements, managing the schedule, checking deliverables, and paying.
Solution
LinKross (“Link + Cross”) connects clients with overseas freelancers and manages the whole collaboration in one structure: select → agree → manage → verify → pay.

Key features:
- LLM-generated milestones and Definitions of Done (DoD): natural-language requirements become milestones and verifiable completion conditions.
- RAG-based English specifications: a glossary-based RAG translates Korean requirements into English specs that keep Korean IT context.
- Semi-automated deliverable verification: submitted code (GitHub commit SHA) is run in a Vercel Sandbox and checked with Playwright against the agreed DoD.
- Milestone-based invoicing and payment records, with test-net USDC payments.

Results
Measured on three sample projects:
- 91.4% of requirements (53/58) were converted into verifiable DoD conditions.
- 0 omissions of core conditions in English translation (37/37 preserved).

My contribution
- Client landing page and side-by-side comparison of applicants
- Freelancer dashboard and marketplace
- English localization of freelancer-facing screens
- Overall UI refinement and logo design
Tech stack: Next.js, TypeScript, Gemini, Supabase (PostgreSQL), GitHub App, Vercel Sandbox, Playwright, Base Sepolia (test-net USDC)
