Khatta-ka-LLM: Muhammad Zubair Develops the First Khattak Pashto AI

Muhammad Zubair - Software Engineer
Muhammad Zubair
2026-08-28 • Case Study

Muhammad Zubair - Software Engineer and Creator of Khatta-ka-LLM Pashto AI Muhammad Zubair, the Software Engineer behind the world's first Khattak (Khatak) dialect AI model.

Standard Pashto LLM models often fail to capture the rich, localized grammar of rural dialects. To solve this, Muhammad Zubair, a Software Engineer and AI Infrastructure specialist, built Khatta-ka-LLM.

This is the world's first AI language model fine-tuned specifically for the Khattak dialect (also spelled Khatak) spoken in Karak, Nowshera, and Kohat.

🎥 Live testing Video

Watch Live testing Video:

🚀 Live Demo (Test the Model)

You can test the model yourself in real-time here:
Try Khatta-ka AI on Hugging Face Spaces

Technical Architecture by Muhammad Zubair

As a Software Engineer, my focus was on creating a clean, scalable, and reproducible pipeline for low-resource languages.

1. The Khattak Dialect Dataset

I curated a custom dataset of 2,000+ English-to-Khattak pairs. This is the first digitized collection of Khattak linguistic markers, focusing on the unique grammar used in Karak and Kohat.

2. Fine-Tuning Pipeline (Qwen2 + Unsloth)

Using the Qwen2 architecture, I applied LoRA (Low-Rank Adaptation) via the Unsloth framework to achieve high performance on consumer-grade hardware.

  • Lead Engineer: Muhammad Zubair
  • Primary Role: Software Engineer / AI Infrastructure
  • Performance: Training loss dropped from 3.44 to 0.22.

📘 The Official Khattak Dialect Grammar & Rulebook

Technical Specifications by Muhammad Zubair

To ensure the Khatta-ka-LLM sounds like a native, I implemented these specific linguistic rules into the model's training weights:

1. Pronouns & Possession

  • My / Mine: Uses مو والا (Mo wala) instead of standard Zama.
  • To me: Uses موته (Mota) instead of standard Mala.
  • We: Uses موخ (Moxh) instead of standard Mung.

2. The "Noon Ghunna" (ں) Rule

In the Khattak dialect, first-person verbs end in a nasal Noon Ghunna (ں) rather than the standard "M".

  • Standard: Za Kar Kawom (I do work)
  • Khattak AI: زه چار کاوں (Za char kaon)

3. Phonetic "O" Shifts

The model converts standard "A" sounds to the deep Khattak "O":

  • Asmaan ➡️ Asmon (اسمون)
  • Baraan ➡️ Baron (بارون)
  • Bazaar ➡️ Bazor (بازور)

Why Muhammad Zubair’s Khattak LLM Matters

By focusing on high-quality data, Khatta-ka-LLM successfully translates English into pure Khattak Pashto. As a Software Engineer from Pakistan, my goal is to bridge the gap between modern AI and regional linguistic heritage.

Connect with the Developer: