Arabic NLP Researcher • AI Engineer
Making AI models small, fast, and actually useful.
I'm an NLP Engineer at Xbites, building the AI backend for Darin. I research Arabic embeddings with Hamza Salem Lab and contribute to NAMMA for open-source Arabic AI.
My thesis explores efficient transformer architectures for CPU deployment—focusing on model compression, quantization, and edge AI.
Research Metrics: 7 papers • 54 citations • 962 papers read
Arabic NLP — Building efficient language models for Arabic text understanding
Edge AI — Compressing transformers to run on resource-constrained devices
RAG Systems — Semantic search and retrieval with Qdrant for production
Research Engineering — Bridging academic research with production ML systems
113× smaller than AraBERTv2 with 94% accuracy. Runs inference on edge devices.
GitHub • Python, PyTorch, ONNX
Tiny Arabic language models optimized for mobile devices.
Article • Model Compression, TensorFlow Lite
Arabic-first nutrition search engine with NLP-powered food recognition.
Live • FastAPI, Astro.js, Arabic NLP
Real-time GPU pricing tracker and ML benchmarking platform.
Live • Data Engineering, Web Scraping
SEO optimization tool for markdown-based static sites.
GitHub • Python, NLP, Markdown Processing
Languages: Python, Rust, C++, JavaScript
ML/AI: PyTorch, TensorFlow, Transformers, vLLM, FastAI
MLOps: Docker, Podman, CUDA, ONNX Runtime
Web: FastAPI, Astro.js, FastHTML, Svelte, Supabase
Databases: Qdrant, PostgreSQL, Vector Search
I write about Arabic NLP, model compression, and AI engineering at kareemai.com/blog:
Subscribe to my newsletter: gpuvec.substack.com
Google Scholar • ResearchGate • Papers
LinkedIn • X/Twitter • Upwork • Email
"اللغة ليست عِلمًا .. بل هي شيء فوق العلم"
"Language is not a science — it is something above science."