Since AI is a Finland-based nonprofit advancing practical AI research, open collaboration, and real-world innovation. We bring together builders, researchers, and organizations to develop and evaluate AI that solves meaningful problems.
We publish models, datasets and tools that are small enough to run anywhere, evaluated on held-out data, and documented well enough to be reproduced.
Research
EuroTender — CPV classification for EU public procurement
Suggests Common Procurement Vocabulary codes for tender notices in any EU language, on a CPU. Trained and evaluated on TED notices with a leakage-controlled temporal split.
EuroDocBench — key-field extraction from European invoices
A synthetic, exactly labelled benchmark in 12 locales and 5 layouts, with a born-digital text track and a simulated-scan track. Tests the details that break document AI in Europe: number and date formats, VAT breakdowns, local labels.
EuroPII Guard — European personal-data detection and redaction
Finds national identity numbers of 22 countries, EU VAT numbers, IBANs, payment cards, phone numbers and e-mail addresses with checksum validation, locally and in the browser. A privacy-assistance tool, not a compliance guarantee.