Skip to content

AI quantity take-off pipeline

Construction drawings in, quantities out, for an estimating firm.

Freelance, sole developer · Construction estimating firm · 2025

37
test modules

How it is built, step by step

  1. Step 1

    Capture

    Playwright captures every drawing sheet.

  2. Step 2

    Route and read

    Each sheet type goes to the right Claude vision extraction.

  3. Step 3

    Review

    Estimators review quantities in a Flask dashboard.

  4. Step 4

    Never regress

    A golden-file harness diffs every change against known-good take-offs.

AI quantity take-off pipeline: system sketchquantitiesDrawingsPlaywright capturePer-sheet routerClaude visionFlask dashboardGolden-file harness37 test modules

In detail

I took a prototype to production: drawing capture, per-sheet vision extraction, a job dashboard, and a golden-file regression harness.

Estimators spend hours reading construction drawings to count quantities. This pipeline captures the drawings with Playwright, routes each sheet type to Claude vision extraction, and presents the results in a Flask job dashboard.

It is covered by 37 test modules and a golden-file regression harness, so a change to a prompt or a parser shows up as a diff against known-good take-offs before it reaches an estimator.

Next: Donation-management system

Write to me about a role

umer.jahangier@gmail.com

Subject: Role: let's talk

A line about the team and the role is enough; I reply within a day.

Open your mail app