/ Case studies / Trexima / Company verification automation

An AI tool for verifying public register data

An internal tool for TREXIMA checking de minimis eligibility across six registries in parallel, returning a decision with an evidence report.

ClientTREXIMA Bratislava, spol. s r.o.
Year2026
Project typeWeb application
AI tool for automated verification of public register data — ukážka rozhrania 1
AI tool for automated verification of public register data — ukážka rozhrania 2
About the client

TREXIMA Bratislava, spol. s r.o. is a Slovak company operating in labor market research and compensation. Verifier is its internal tool for administering the DM-08/2024 minimal-aid scheme — an unpublished, login-protected application for internal use.

Starting point

The goal was to automate the decision process for verifying a company's eligibility for de minimis aid under the DM-08/2024 scheme — combining data from six different registries in real time without manually searching each one separately, while fully respecting GDPR.

Challenges

Building this compliance tool, we tackled several key challenges.

  • Six sources, one consistent picture

    Financial, ownership and personnel data are scattered across FinStat, ORSR, RPO, RÚZ, Transparex and IS SEMP — these had to be merged into one consistent case.

  • Aid is calculated per group, not per company

    De minimis rules apply to a "single undertaking" — the entire network of companies connected through ownership — the tool had to reconstruct ownership graphs and correctly sum aid across the whole group.

  • Real data with no permanent storage

    A GDPR-stateless architecture required that the verified entity's data never be saved to a database — only a temporary cache and a one-time downloadable report.

  • Runs of minutes, not seconds

    Some checks take several minutes for larger networks of connected companies — progress had to be shown live instead of leaving an empty wait.

Our approach

We built a Next.js application with login, company lookup by registration number or name, and a verification run that pulls data from all six sources in parallel, streaming progress live via Server-Sent Events.

  • Company lookup by registration number or name with autocomplete via FinStat
  • Parallel data retrieval from 6 sources (FinStat, ORSR, RPO, RÚZ, Transparex, IS SEMP)
  • A live progress indicator for the verification run via SSE
  • Automatic VIABLE / NOT_VIABLE / NEEDS_REVIEW decisions with reasons
  • A complete DOCX evidence report with an ownership diagram and a de minimis aid table
  • GDPR-stateless mode — the report is deleted immediately after the first download or after one hour
Process
1

Registry and rule analysis

We mapped six external data sources and the rules for calculating de minimis aid across connected companies.

2

GDPR-stateless architecture design

We designed a system with no permanent storage of the verified entity, only a temporary TTL cache.

3

Development and parallel scraping

We implemented a Next.js application with parallel retrieval from 6 registries via SSE and Playwright scraping of Transparex.

4

Decision logic and reporting

We built the evaluation of NACE exclusions, the aid ceiling and size category, with generation of a DOCX evidence report.

Results

Verifier automated a decision process that previously required manually searching six registries at once — a reviewer now enters a registration number and gets a clear decision within minutes, with an exportable evidence document.

INTERFASE