VergabeHero started from a simple problem: public tenders in Germany are scattered across dozens of portals, and the companies that should win them lose hours every week to manual research. The initial goal was to build an autonomous system that continuously crawls every relevant procurement source — Bund.de, eVergabe, DTVP, Vergabe24, ANKÖ and TED — and surfaces only the announcements that actually fit a given company.
I built the AI pipeline end to end: crawlers that import and deduplicate thousands of new tenders every hour, an extraction layer that turns a company's website, past bids and PDFs into a structured "digital twin" of its services, certifications, regions and references, and a semantic matching engine that scores each tender against that profile with a transparent, per-factor breakdown. At peak the system processes 14,200+ new tenders a month and takes a company from announcement to ready-to-submit bid in one place.