Anastasiia Koloskova (EPFL 2026)
Much of society’s most valuable data — including medical records, financial transactions and environmental observations — is too sensitive or too widely dispersed to be brought together in a single location. This limits the use of modern artificial intelligence methods, despite the major advances they have enabled in many other fields.
This project seeks to address this challenge by bringing together three essential objectives: protecting data privacy, enabling collaboration among a large number of participants, and training artificial intelligence models accurately and efficiently.
By developing algorithms designed for environments with limited resources and strict privacy requirements, the project will enable hospitals, universities and smaller institutions to benefit from advances in artificial intelligence without having to centralise sensitive data or rely on large-scale computing infrastructure.
Ultimately, the project will help make artificial intelligence more trustworthy, sustainable and widely accessible.