Skip to content
View Kagankakao's full-sized avatar
☕
Turning coffees into code.
☕
Turning coffees into code.
9 followers · 0 following

Achievements

Achievement: YOLOAchievement: Pull Shark

Achievements

Achievement: YOLOAchievement: Pull Shark

Highlights

  • Pro

Block or report Kagankakao

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don't include any personal information such as legal names or email addresses. Markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Kagankakao/README.md

Banner

Kağan Arıbaş

AI / NLP Engineer • SE undergrad @ Fırat University • Turkish morphology & tokenization • Production RAG and retrieval optimization

Portfolio

Visit my portfolio for demos, case studies, and deeper project details: https://aribaskagan.up.railway.app/ (Best viewed on desktop.)

I build practical NLP systems with a focus on Turkish language modeling, efficient retrieval, and LLM app quality. Recently I am working on FST-based morphology, RAG pipelines, and evaluation-driven improvements.

Focus

  • RAG systems: retrieval, chunking, re-ranking, and evaluation loops
  • Turkish NLP: morphology, tokenizers, and linguistic pipelines
  • LLM applications: chatbot integration, cost/latency optimization, and tooling

Highlights

  • Reduced token usage and API cost by ~80–90% with system-level optimizations while preserving quality
  • Built an FST-based Turkish morphology model with 65K+ stems and 200+ suffix forms
  • Designed a Viterbi-based POS disambiguation approach for context-sensitive analyses

Projects

Tech

  • Languages: Python, C#, C++, Java
  • ML/NLP: PyTorch, scikit-learn, Transformers, embeddings, RAG
  • Data: Pandas, NumPy, Matplotlib
  • Infra: Docker, Linux, Git

Collaboration

  • Problem solving: structured approach, clear trade-offs, and measurable outcomes
  • Teamwork: dependable in cross-functional teams, open to feedback, and supportive in reviews

Links

Stats

Pinned Loading

  1. IpekYoluGPT/ipekgpt IpekYoluGPT/ipekgpt Public

    IpekGPT is a RAG-powered assistant built for the İpek Yolu Uluslararası Çocuk ve Gençlik Çalışmaları Merkezi. It provides accurate, organization-specific answers by combining a curated knowledge ba…

    Jupyter Notebook

  2. TurkishTokenizer/turkish-morphological-segmentation TurkishTokenizer/turkish-morphological-segmentation Public

    Research-grade Turkish morphological segmentation system and dataset (roots, suffixes, POS) built from Kaikki, Zemberek, and Wikimedia; optimized for FST-based linguistics.

    Jupyter Notebook

  3. KEGOMODORO KEGOMODORO Public

    KEGOMODORO is a Pomodoro timer app designed to infuse your work and break cycles with creative energy. Enjoy a fully customizable countdown that adapts to your unique workflow. Integrated with Pixe…

    Python 13 2

  4. FitTurkAI FitTurkAI Public

    Forked from FitTurkAI/FitTurkAI

    AI-powered fitness assistant that generates personalized workout and nutrition plans, adapts to progress, and supports goal-driven coaching with smart recommendations.

    TypeScript

  5. personal-os personal-os Public

    RAG-powered AI productivity dashboard with built-in notes/journaling plus KEGOMODORO and Pixe.la integrations.

    Python

  6. ViT-from-Scratch ViT-from-Scratch Public

    A repository focused on the development and exploration of AI & ML techniques, featuring projects, code, and resources that document my learning journey.

    Jupyter Notebook