Product & systems deliveryJan. 2025–present

B2B Digital Operations Platform

Public scope

This case study uses information already disclosed in the public CV. It omits proprietary business rules, credentials, customer data, internal records, and source code.

01Product surfaceMini Program, back office, website
02OperationsFive connected business domains
03Application34-page Mini Program
04OwnershipRequirements through adoption
Methods & tools
uni-appVue 3TypeScriptFastifyPrismaPostgreSQLRedisPM2Nginx
01 / Context

One operating model across fragmented workflows

The platform connects customer-facing and internal operations that previously depended on separate processes. My role spans product definition, workflow design, data architecture, technical implementation, deployment, and adoption rather than a single functional layer.

02 / Product scope

Three interfaces serving one system

The product surface includes a WeChat Mini Program, a back-office system, and a company website. I led the delivery of a 34-page uni-app, Vue 3, and TypeScript Mini Program alongside the supporting backend and operational interfaces.

  • Customer and partner workflows through the WeChat Mini Program
  • Operational administration through the back-office system
  • Company information and external access through the website
03 / Data & workflow

Five operational domains, one data architecture

I designed workflows and shared data structures across inventory, procurement, production, sales, and finance. The work required translating operating practices into product requirements, states, permissions, records, and acceptance criteria that could be implemented consistently.

04 / Engineering delivery

From product requirements to production deployment

The backend uses Fastify, Prisma, PostgreSQL, and Redis, with PM2 and Nginx supporting deployment. I own the full lifecycle from requirements and architecture through implementation coordination, testing, release, and user adoption.

05 / Relevance

Why operational systems matter to research

Building a production system makes control design concrete: requirements must map to data, workflows, permissions, interfaces, tests, and operational behavior. That experience informs how I think about turning evaluation findings into controls that teams can actually implement and verify.

Open channel / Collaboration

Working on agent safety,
evaluation, or multilingual AI?