Katja Gorlinski

Founder of iEVA (Finland), a platform that trains decision-making under pressure. The project developed from the iEVA Method and seven years of direct work with founders, managers, teams, and individuals. I came to AI safety through this work. In 2019, I formulated a working hypothesis that the quality of data and environment loaded into a decision-making agent, whether a person learning cognitive skills or a model in training, essentially determines the safety of the systems that agent operates. Years passed and the whole field turned to sociotechnical alignment. My hypothesis found itself in good company. This fall, I am piloting the human part with a real company team: baseline, behavioral metrics, pre-registered endpoints. This project examines the machine part of the same hypothesis: does curated data change the dispositions of a model? I build the project solo, in the EU, and I believe in claims that can be falsified.

Personal Website

https://www.i-eva.fi/

LinkedIn

LinkedIn

Projects

Does the thinking behind data selection shape a model's character?

Team Member

No grants recorded.