About
⚡ TL;DR
- 🧠 ML Engineer focused on RAG, Computer Vision, and Arabic NLP
- ⚙️ Build and run AI systems locally (RTX 4050, no API reliance)
- 🚀 Creator of TaLibAI — bilingual Arabic-English RAG for technical docs
- 🏗️ Built and deployed full MLOps pipeline (Docker, CI/CD, Azure)
- 🔬 Strong focus on model behavior, ablation studies, and real-world reliability
╔══════════════════════════════════════════════════════════════════╗
║ DRAGOX7 SYSTEM — BOOT LOG v2026.04.12 ║
╚══════════════════════════════════════════════════════════════════╝
[ OK ] Loaded: NVIDIA GeForce RTX 4050 Laptop GPU .............. ✔
[ OK ] Loaded: PyTorch 2.x + CUDA ............................ ✔
[ OK ] Loaded: TaLibAI RAG Pipeline (Gemma 4) ................ ✔ [80%]
[ OK ] Loaded: ZenML MLOps Orchestrator ...................... ✔
[ WAIT ] Loading: Unity 2D Action RPG Project .................. ⟳
[ OK ] ENV: UNIVERSITY="Jordan University of Science & Technology"
[ OK ] ENV: FOCUS="Computer Vision · MLOps · Arabic NLP"
[ OK ] ENV: PHILOSOPHY="Run it locally. Own the pipeline."
root@DRAGOX7:~$ _
👤 About Me
I'm a 3rd-year Computer Science student at JUST, Jordan who doesn't wait for cloud credits. My RTX 4050 runs experiments at 3 AM. I read papers and then replicate them — parameter counts included. My work sits at the intersection of research depth and production discipline: ablation studies, MLOps pipelines, RAG architectures, and real inference on real hardware.
I care about building AI that works in the real world — especially for Arabic and technical Arabic contexts that most English-first tools ignore.
"Most people use models. I study why they break."
🚀 Current Research & Focus
| Process | Status | Priority |
|---|---|---|
| 🤖 TaLibAI — RAG pipeline (Gemma 4 + local embeddings) | [▓▓▓▓▓▓▓▓░░] 80% | HIGH |
| 👾 Unity 2D Action RPG — AI-integrated game mechanics | [▓▓▓░░░░░░░] 30% | MEDIUM |
| 🔬 Retrieval Latency Optimization — TaLibAI sub-module | [▓▓▓▓▓░░░░░] 50% | HIGH |
| 📚 Gemma 4 Fine-tuning — domain adaptation | [▓▓░░░░░░░░] 20% | ACTIVE |
📂 Featured Experiments & Projects
🤖 TaLibAI — The Main Event
RAG Pipeline for Technical Documentation | Python · Gemma 4 · PyTorch · Local Embeddings
Not another LLM wrapper. TaLibAI is a ground-up retrieval-augmented generation pipeline designed to reason over dense technical PDFs (C++, ML textbooks). Built with custom chunking strategies tuned for code-heavy documents, local vector embeddings for full data privacy, and optimized query latency on consumer hardware.
Architecture: PDF Ingestion → Chunker → Local Embedder → Vector Store → Gemma 4 → Response
Hardware: NVIDIA RTX 4050 (no API bills)
St
Technology
Selected Work
A Deep Learning application (DenseNet121) for detecting chest X-ray abnormalities. Built with PyTorch & FastAPI, containerized with Docker, and automatically deployed to Azure via GitHub Actions (CI/CD).
TaLibAI: A Multilingual RAG Study Assistant for JUST University students, featuring local PDF/Image ingestion, FastAPI background processing, and Gemini/Ollama integration.
End-to-end MLOps pipeline and NLP engine analyzing Levantine/Jordanian Arabic sentiment using a custom PEFT/LoRA architecture (MARBERTv2).
Low-light image enhancement using Retinex, CLAHE, and Deep Learning techniques.