Hi there! 👋 I create niche, high-quality datasets to help Turkish Large Language Models (LLMs) speak more naturally, fluently, and in tune with local dynamics. My goal is to train AI not just with "translated Turkish," but with the real Turkish spoken on the streets, in the office, and in everyday life.

🔥 Featured Work

🕵️ Kusursuz Ajan (Flawless Agent): Turkish RLHF & DPO Dataset

Specifically designed to teach AI models human nuances like manipulation, sarcasm, passive-aggressiveness, and street smarts through challenging dialogue scenarios.

DPO & RLHF Ready: Structured with prompt, chosen, and rejected pairs, ready for direct training.
Cultural Accuracy: 50+ Turkey-specific categories covering family, corporate life, bureaucracy, shopping, and even taxi drivers.
Natural "Dirty" Text: Features organic inputs with typical internet laziness (missing letters, lack of punctuation, lowercase usage) for robust training.

👉 Check out the Full Version on Gumroad https://abidin1211.gumroad.com/l/oapgbr

🛠️ Core Focus Areas

RLHF & DPO: Generating data to align models with human preferences.
Cultural Context: Crafting dialogue trees that reflect Turkey's local dynamics, avoiding the "stiff translation" feel.
Red Teaming & Safety: Challenging prompt engineering to test how models react to toxic, manipulative, and demanding user inputs.

📬 Contact & Support

💌 Email: abidin.esen@hotmail.com 🛒 Gumroad: https://abidin1211.gumroad.com/

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support