Tomas Bujnak.

Reveal

Are the datasets you're training your AI on actually anonymous? In 2000, Latanya Sweeney showed that date of birth, ZIP code, and gender alone uniquely identify 87% of the US population. Three columns. No names required. By 2019, Rocher et al. confirmed that 99.98% of individuals remain re-identifiable from 15 demographic attributes. Three studies, two decades, one finding: removing names is not anonymization.

Reveal reads only your CSV column headers, never the rows. It calculates your k-anonymity score, identifies re-identification risks, maps GDPR exposure, and tells you exactly what to change. Not a legal audit. A fast first check before you talk to your DPO. Built in 3 hours at Give(a)Go Builder Day.

VSZK Systém

Central tournament management system I built on years of experience as a competitor, coach, and organizer. Designed for every role: organizers, referees, competitors, parents, and spectators. Live scoring, bracket generation, timetable logging, and a public spectator page with real-time results and delays. Handling dozens of teams, 50+ categories, and 300+ individual starts per tournament.

Copilot Inšpirácie s AI

Six-week AI competition for Slovak educators. I work on task design, scoring logic, weekly comms, and the social media push. A hands-on program that turns AI literacy into an outcome teachers actually take back into the classroom.