2025
Schematic Verification Platform
Full-stack tool for Pörner GmbH that cross-checks building schematics against equipment lists and flags missing or corrupted documents, replacing a manual verification process with a custom OCR pipeline.
The problem
Engineering plans and the records that describe them, things like equipment lists and asset registers, are supposed to agree with each other. They drift apart instead, and catching that meant an engineer reading large format PDF schematics against a spreadsheet by hand.
It was worst when onboarding a new client whose plans only existed as printed documents, with no structured data behind them at all.
The approach
A custom OCR pipeline reads the schematics, pulls out the entities on them and compares those against the related records. Anything missing, unreadable or inconsistent gets flagged for review instead of quietly passing.
A Flask service does the parsing and comparison, and staff work through the results in a Next.js frontend. It is containerised so it can run inside the client's own environment.
The outcome
A document by document manual check became an automated pass that only surfaces the discrepancies, which is what made scanned-only plan sets practical to take on at all.