Research
The research behind the product
Within Selçuk University Technopark we run applied research that feeds our products — not for academic display, but for accuracy measured in the field.
- Technopark R&D status
- Findings published as technical notes on the blog
01Focus areas
Where we go deep
Turkish legal NLP
Domain entity recognition, citation extraction and decision summarisation — developed and measured on a corpus of 11M+ rulings.
Citation verification
Automatic matching of every generated citation against the original source — hallucination prevented at the architectural level.
Personal-data masking
High-precision NER-based masking and irreversible pseudonymisation for Turkish text; the research layer of a gateway running in production.
RAG evaluation
Domain-specific measurement of groundedness, hit rate and answer quality; evaluation sets built from real professional questions.
02Technical notes
Latest writing
Strategy ·
Why an Enterprise AI Project When ChatGPT Exists?
Why invest in enterprise AI when ChatGPT exists? An honest comparison across data privacy, citations, access control, process integration and cost.
2 min read · Read →Architecture ·
KVKK-Compliant LLM Architecture: The Mask-Then-Process Chain
How to build a KVKK/GDPR-minded LLM architecture: the production-proven chain where personal data is masked before any model call.
2 min read · Read →Fundamentals ·
What Is RAG? Enterprise Knowledge Assistants, Plainly Explained
What is RAG (Retrieval-Augmented Generation)? The architecture behind enterprise knowledge assistants, explained plainly with real examples and honest limits.
2 min read · Read →Research collaboration
We are open to collaboration with university groups and researchers on Turkish NLP and trustworthy AI.