←All work02 / 08
AI for public tenders
Helps companies win US public tenders — AI-estimated prices that win without bidding below cost.
- Year
- 2025–26
- Role
- Full-stack developer in a team of 7
- Client under NDA
- Stack
- .NET · Angular · Rust · PostgreSQL · Local LLMs · Jev (TypeSafe AI) · OCR
Overview
A platform for companies bidding on public tenders across the United States. It imports tenders from documents of any format with OCR, scores the chance of winning, forecasts upcoming tenders and suggests prices that win the contract while keeping losses to a minimum. Historical tenders and analytics are broken down by state.
In numbers
- AI layers: OCR · LLMs · Jev
- 3
- engineers on the team
- 7
What I built
- 01Estimates as draft bids: turning a tender into an editable estimate.
- 02LLM-generated pricing for individual tender line items.
- 03An admin panel for managing historical tenders.
- 04Market Insights — a module where LLMs generate insights for each region.
- 05The product's landing page, plus code review across the team.
Highlights
- OCR import of tender documents in any format.
- Local LLMs for generation and Jev (TypeSafe AI) for typed, calibrated decisions.
- .NET and Angular, with Rust microservices on PostgreSQL.