Eschatia Labs labs

About

A small collective of researchers.

Eschatia Labs is an independent, non-profit open-source AI research lab. Six of us work together as a collective: we collect our preprints on this site, publish our models on Hugging Face and our code on GitHub, and share all of it with everyone.

We build small models for specific jobs and the open tests that measure them: decision models for English, Turkish and German, local task assistants, explainable deepfake detection and a model for Tolkien's Elvish languages. Eschatia is Ancient Greek for the outer edge of a territory.

eschatia ἐσχατιά the outer edge

How we work

  • Each model is trained for one job and tested on that job.
  • Test questions stay out of the final training run.
  • Results, known failures and the files to repeat them are published together.
  • The released JevAlt files run on a laptop processor with about 3 GB of memory.

Founders

  • Mert Kaya, founder. Senior AI engineer at Enocta. He works on explainable deepfake detection, small decision and language models, local task assistants and tools that coordinate coding assistants. MSc in Computer Science, TED University.
  • Betül Keleş Kaya, co-founder and head of design. Master's student in architecture at Anhalt University of Applied Sciences (DIA); graduated top of the architecture class at Çankaya University.

Image credits

View image credits Sources and licences for the artwork used on this site.

EN · TR · DE · ZH · IT · ES · Last updated November 7, 2026