Skip to content
→ All work

Case study 08 / 26

DataConstruct

A manufacturing platform built around one loop — plan, actual, variance, why, action, result — with finite-capacity scheduling and tenant isolation enforced by PostgreSQL row-level security.

Status
In development
Domain
web · cloud · security
Source of claims
Private repository README (reviewed). Source and deployment are not public.

ERP, MRP, MES and quality management for manufacturers: BOMs, routings, yield, costing, capacity and variance, with a drag-and-drop Gantt scheduler, offline operator mode and a grounded assistant.

01/The problem

Factories know what should happen and, eventually, what did — but rarely why the two differ. DataConstruct is built around closing that loop on the shop floor.

02/The system

ERP, MRP, MES and quality management for manufacturers: BOMs, routings, yield, costing, capacity and variance, with a drag-and-drop Gantt scheduler, offline operator mode and a grounded assistant.

03/Scope

  1. 01Pure calculation engines for BOM, yield, cost, variance, MRP, capacity and root cause, in Decimal maths.
  2. 02Production scheduling with finite capacity, conflict detection and a drag-and-drop Gantt.
  3. 03Quality, downtime and CAPA workflows; offline operator mode with documented sync and conflict rules.
  4. 04A demo company loaded by driving the real API, so every figure comes from the same workflows a user runs.

04/Engineering

Isolation the database enforces

Tenant requests run through a role subject to PostgreSQL row-level security, and the API refuses to start if that role could bypass it.

Migrations that keep their guards

Append-only triggers and frozen standards live in hand-reviewed migrations; tests create, migrate and drop their own uniquely named database.

05/Interface

Interface screenshots of this commercial product are not public. The visual above is an abstract representation of its modules — not the product itself.

06/Tech stack

  • NestJS
  • Next.js
  • TypeScript
  • Prisma
  • PostgreSQL (row-level security)
  • Redis
  • BullMQ
  • Decimal maths

07/Result

Verified outcomes

  • 149 unit tests for the calculation engines, plus API integration tests against real PostgreSQL.

Known limitations

  • Not yet ready for real customer data — onboarding, backups and production providers remain open.

08/Links

Private commercial codebase — no public links.

Next case study

DataItemize →