← back to portfolio

Production Cost Calculator AI

AI Computer Vision Cost Modeling

An AI system that reads technical production drawings and generates accurate cost estimates in seconds. Replaced a multi-hour manual process.

The problem

Every production job required someone to manually read technical drawings, identify materials, measure dimensions, calculate quantities, and look up pricing tables. A single estimate could take hours. Multiple estimates per day meant the person doing this was bottlenecked permanently.

The approach

I built an AI system that interprets uploaded technical drawings, extracts dimensions and material specifications, and cross-references them against production cost tables to generate an estimate automatically.

DRAWING pdf / dxf VISION parse + extract MATERIALS classify COST MODEL live pricing ESTIMATE + reasoning human review on every output
System pipeline — drawing to estimate, with human-in-the-loop verification

The outcome

What used to take hours now takes seconds. The system isn't perfect, but it's right often enough that the human review step takes minutes instead of starting from scratch. Three months after deployment, nobody mentions the tool anymore. It's just how estimates work now. That's the goal.

~95%
Material ID accuracy
<10s
Time per estimate
180×
Speedup vs. manual
PRODUCTION ESTIMATE · JOB #4280-A SARA NOOR DRAWING 1240mm 820mm 820mm A-A B-B EXTRACTED Material MDF 18mm Surface area 3.42 m² Edge length 7.81 m Hardware 4 hinges, 2 handles Finish matte lacquer COST BREAKDOWN Material AED 312 Hardware AED 88 Labour AED 240 Finishing AED 96 TOTAL AED 736
Interface representation — drawing parser with extracted spec and cost breakdown

What I learned

The temptation was to hide the AI's reasoning and just show the number. Faster, cleaner. But that's how you build systems people stop trusting the moment they get one wrong answer. Showing the work, the dimensions it extracted, the materials it assumed, keeps the human in the loop without slowing them down.