Bicitra EMS
Product Architect / Lead Engineer
A multi-tenant Education Management System for schools, madrasas, and colleges: one shared platform serving any number of independently isolated institutions, built to run on nothing more than a phone: QR-code student IDs, camera-based scanning, and a bilingual (Bangla/English) interface instead of dedicated hardware.
- Client
- Schools, madrasas and colleges in Bangladesh. My own product.
- Problem
- Schools run on paper registers, and setting up software one school at a time costs too much.
- What I built
- One platform many schools share, with each school's data kept apart: web dashboards, and a mobile app for parents, teachers, gate staff and bus drivers. QR attendance, fees, records and live bus tracking.
- My role
- Product Architect / Lead Engineer
- Timeline
- About three weeks of solo building. Heading to launch.
- Stack
- NestJS, TypeScript, Prisma, PostgreSQL, Next.js, Flutter, Redis, Docker
Problem
Schools in Bangladesh run on paper registers (admission, attendance, fees, stock, visitors, meeting minutes) that are disconnected from parents and expensive to digitize one institution at a time. Bicitra EMS is built as a single multi-tenant platform instead: any school signs up, self-serves onboarding, and runs on shared infrastructure with its data fully isolated from every other school, so the cost of serving the next institution is close to zero. The non-negotiable constraint: it has to work on a single Android phone per role (teacher, gateman, bus driver), with no dedicated scanner hardware, QR codes plus a phone camera standing in for all identity verification.
Approach
A monorepo with three apps sharing one domain model: a NestJS/TypeScript API on Prisma and PostgreSQL, a Next.js web dashboard for institution and platform admins, and a Flutter app covering four personas (parent, teacher, gateman, driver). Multi-tenancy is enforced structurally rather than by convention: every tenant table carries a schoolId, and isolation is applied by a Prisma Client Extension in apps/api/src/prisma/ so a query can't accidentally skip the scope; a caller doesn't get the chance to forget a WHERE clause. A separate platform-admin console handles cross-tenant billing, plan enforcement, and analytics, reusing the same tenant-scoped services as the school-facing app rather than duplicating query logic. Built in phases (admission and QR identity first, then fee collection and academic records, then transport tracking and the parent portal, then bilingual UI and analytics), with push-to-pay, SMS delivery, and native push notifications deliberately built to a swap-later boundary: the production shape is in place, waiting on real payment-gateway and SMS-vendor credentials rather than on more engineering.
A hard problem I solved
I put 338 automated checks behind the product (225 API tests against a real Postgres, 66 Playwright browser tests against a real build, 47 Flutter tests) specifically because tenant isolation and money arithmetic are exactly the two things a mock can't tell you anything about. Writing them wasn't a formality: they found six real defects in code that had already been called done, the worst of which was a critical cross-tenant data leak. The isolation extension correctly scopes every row it writes, but a nested Prisma include on a relation is evaluated as part of its parent query and never gets routed back through that extension. So a fee invoice, an attendance record, or a salary row referencing another school's student id would read that school's guardian details and QR gate-scan credential straight back out to the wrong tenant. It was reachable through seven different endpoints. Rather than patch each one, I fixed it structurally: every inbound tenant-scoped foreign key is now declared in one central registry and validated at write time inside the same extension that already owns tenant scoping, so a new endpoint gets the protection automatically instead of depending on whoever writes it remembering to add it. The same test pass also caught a double-payment race condition (two simultaneous 'Pay now' taps on a flaky connection could both record a payment) and a gate console accidentally handing staff password hashes to the lowest-privilege role in the building.
Status
Currently building toward launch, after roughly three weeks of solo full-stack engineering across the API, web, and mobile apps, with a collaborator (Hossain al Ashrafi) contributing the registers and government-compliance requirements. Core registers, QR identity, class and gate attendance, fee invoicing, academic records, live bus-position tracking, and a full bilingual Bangla/English interface are built and covered by the test suite above. Real mobile-banking (bKash/Nagad/Rocket) integration, SMS delivery, and device push notifications are the remaining pieces, blocked on vendor credentials, not on unbuilt architecture.
Technology
- NestJS
- TypeScript
- Prisma
- PostgreSQL
- Next.js
- Flutter
- Redis
- Docker
Writeup
Hiring a developer? Get the checklist first
20 checks founders can run before they sign: questions to ask on the first call, who should own the code and accounts, and what senior remote developers cost in 2026. A free two-page PDF, plus three short emails from real builds. Unsubscribe any time.
Need something similar built?
I build and run products like this for founders and businesses worldwide, from architecture to deployment.
Or tell me about your project →Related: SaaS Development for Startups, Next.js, Node.js & NestJS Development, Fractional CTO, Architecture Reviews and Code Rescue