Building NLP systems for the 400M+ Arabic speakers who deserve better language tools.

Currently: RAG architecture for Arabic Q&A • Training competitive programmers • CS at ENSIA

Arabic NLP that actually works
Most language models treat Arabic as an afterthought. I'm building RAG systems, tokenizers, and Q&A pipelines specifically designed for RTL languages and Arabic morphology.

Production ML, not just notebooks
From medical diagnosis pipelines to security-focused mobile apps, I focus on systems that ship—with proper evaluation, reproducible research, and real impact.

Algorithms education
Coach for Algeria's national programming olympiad. Turns out teaching is the best way to truly understand complexity theory.

Core: Python • C++ • TensorFlow • scikit-learn • Hugging Face
NLP: LangChain • RAG • Tokenization • Arabic morphology
Mobile: Flutter • Dart
Tools: Linux • Git • LaTeX • Jupyter

RAG-powered Q&A system for Arabic government documents. Natural language search instead of ctrl+F through 500-page PDFs.
Python • LangChain • Arabic NLP

Flutter app with ML-based phishing detection. Analyzes QR codes before you scan them, not after you're compromised.
Flutter • ML • Security

End-to-end ML pipeline for medical diagnosis. Focused on reproducibility and proper cross-validation—because healthcare predictions need more than 0.95 accuracy on a test set.
Python • scikit-learn • Medical ML

Interactive demos of pathfinding algorithms. Built this for my olympiad students who learn better by seeing, not memorizing pseudocode.
Python • Algorithms • Teaching

More at github.com/Samir-Guenchi

There are 400 million Arabic speakers online. Most NLP tools were built for English, then poorly adapted. I'm working on infrastructure that treats Arabic as a first-class citizen—proper tokenization for agglutinative morphology, embeddings that understand context in RTL text, RAG systems that handle diacritics.

Also: competitive programming teaches you to think in constraints. That mindset carries over when you're optimizing transformer inference or designing algorithms that scale.

• Building Arabic RAG systems at ENSIA (Algeria's National School of AI)
• Training national olympiad candidates in algorithms
• Contributing to Arabic NLP tooling
• Looking for: Research collaborations, internships in ML/NLP, open-source opportunities

LinkedIn • Email • Kaggle • Codeforces