KO/EN
← Back to projects
liveClient work·2026.01 – 2026.06

Mirlink

Storing dental-implant records on-chain — from MVP to production

Client
Healthcare venture (patented, university-hospital partner)
Team
4 people — Design review 1 · Infra 1 · Support 1 · Me
My role
Frontend, Backend, Smart contract, Infra, POS payment integration

The challenge

Records had to be readable only by the patient — not even by the server — yet elderly patients who lose their card still needed their history back. The challenge was keeping data sovereignty and everyday usability at the same time.

Key decisions

Redesigning around a new payment and settlement flow mid-transition

Moving to production, the payment and settlement model changed wholesale, and with it the hospital and patient user flows. I redrew the flows against the new requirements, assessed what in the existing architecture could stay and what had to change — feasibility and schedule impact first — then agreed scope and priorities with the PM before redesigning.

Key management that survives a lost card

I separated the key that opens the records from the card the patient carries. When a card is lost, a recovery step binds the same key to a new card — and the server never learns the key along the way. Card replacement now works without giving up server-blind storage.

Results

  • Built frontend, backend, contracts, infra and POS integration
  • Tracked progress against feature specs & WBS, shared with the PM
  • Migrated MVP → production on AWS ECS/Fargate + RDS

Tech

  • Next.js — Next.js 16 App Router — hospital-web, patient-web
  • React
  • TypeScript
  • Tailwind CSS
  • TanStack Query
  • Auth (NextAuth · JWT) — NextAuth v5 + JWT
  • Prisma — Prisma 7
  • PostgreSQL — NeonDB → AWS RDS
  • Node.js
  • Encryption architecture — Designed lost-card recovery without the server ever holding the key
  • Solidity — MedicalRecordStore.sol
  • Hardhat — Hardhat 3 + Ignition
  • wagmi · viem
  • Avalanche
  • AWS (ECS · RDS · S3) — ECS/Fargate, ECR, ALB, RDS, S3
  • Docker
  • GitHub Actions
  • Monorepo (Nx) — Nx + npm workspaces

Planning & collaboration

  • Architecture & user flows — Reworked user flows and architecture after the payment model changed
  • Feature specs — Feature-spec-based KPIs for completion tracking
  • WBS & scheduling — WBS as the shared deadline baseline with the PM
  • Decision option design — Framed the hospital payment case as three options
  • Working with PMs — Worked with a dedicated planner — flagged constraints early, asked back on edge cases
  • End-to-end ownership — Owned all engineering in a team of four

Evidence