An AI preconstruction platform that turns a month of bid work into a single day

The situation

For every new project, the client's team reviewed 150+ documents — architectural drawings, specifications and more — by hand to work out the job scopes. That alone took more than a week. Then came finding the right subcontractors, writing and sending proposals one by one, and finally comparing the bids that came back, all manually. Getting a single project out to bid took about a month.

The challenge

• ✗150+ documents per project, from drawings to specifications
• ✗Job scopes buried across hundreds of pages
• ✗Subcontractor matching and outreach done one email at a time
• ✗Bids arriving in different formats, compared by hand
• ✗Every decision still needed a human sign-off

What I built

1. 1Designed and built the entire platform end to end — architecture, backend, frontend, AI and deployment
2. 2Document pipeline: project documents are uploaded and processed in the background with Celery workers, stored securely in MinIO
3. 3AI document review: an OpenAI-powered agent reads every document and extracts the job scopes
4. 4Automatic subcontractor matching based on each scope
5. 5AI-prepared proposals for outreach to every matched subcontractor
6. 6AI bid comparison that ranks incoming bids and recommends the best subcontractor
7. 7Human approval gates at every stage — the AI does the work, a person signs off before anything moves forward
8. 8Award workflow: approve the subcontractor and award the project inside the platform

Architecture

The result

✓Document review that took more than a week now takes about 30 minutes, and a project can go out to bid within a single day instead of about a month. The team keeps full control — every AI step waits for human approval — while the platform does the heavy lifting.

Stack

• Django
• Next.js
• Celery
• PostgreSQL
• Redis
• MinIO
• OpenAI