How did EcoFluxion come about?
EcoFluxion began with a simple question: Why can't a Turkish speaker easily access an AI that is truly powerful and trustworthy in their own language?
“The best models were built for English. So we decided to build one for Turkish.”
It started with a problem
We were following the rise of AI closely. But whenever they encountered Turkish text—especially a legal text—even the most powerful models stumbled: wrong terminology, fabricated sources, answers that missed the context. That's because these models were designed primarily around English. Turkish's agglutinative structure, its distinctive syntax and domain terminology (law, public administration…) were often overlooked.
This was no minor shortcoming; it was a silent disadvantage for millions of Turkish speakers. That is exactly where EcoFluxion's starting point lay.
From research to product
We set out with a lab's curiosity. We researched how we could teach large language models (LLMs) Turkish, and gained hands-on experience training and deploying models on large-scale infrastructure through TÜBİTAK events, the national computing infrastructure TRUBA and programs such as EUMaster4HPC. As the prototypes grew, one thing became clear: the real value wasn't a paper or a demo, but a product people open every day.
2024: Two founders, one company
In 2024, we founded EcoFluxion Teknoloji A.Ş. at ODTÜ Teknokent in Ankara — with two co-founders: İsmail Tarık Şenkal and Halide Yılmaz. We distilled the company's compass into a single sentence: “From data to intelligence, from research to product.”
Why a product, not a service?
Many technology companies do work for others. We chose a different path: building our own products. Because truly solving the Turkish AI problem doesn't happen by delivering a project and walking away; it happens by owning and developing a product for years.
The first output: İçtiHub
The first tangible product of this idea was İçtiHub: a platform that brings together case-law search, document analysis and a smart legal assistant for lawyers and legal departments. Behind it sits our own language model MevzuatBot, which we trained for the Turkish legal language, and a RAG architecture that grounds answers in real documents.
In short, EcoFluxion was born not out of an urge to “let's start a company,” but out of a real problem that needed solving. For the more technical side, you can take a look at our articles on what a large language model is and why Turkish AI.