Architecture and backend first — distributed platforms, real-time pipelines, clean APIs. What I power them with is LLMs, RAG and voice. I don't train models; I make them ship. My specialty is conversational AI that works in Arabic and its dialects, in production, for real enterprises rather than demos.
Agentic voice agents for enterprise contact centers — callers talk to an LLM-driven agent that listens, reasons over enterprise knowledge and answers in real time. Python · FastAPI · MCP · SIP/WebRTC · Milvus/FAISS. +25% response efficiency, −30% operational cost.
Real-time translation on the contact-center floor — 1,000+ multilingual interactions every day, plus an AI assistant that cut average call handling time by 30%.
Languages & frameworks
AI
Data
Infra
The enterprise work lives behind NDAs — these older projects show the Arabic-NLP foundation it stands on.
| Project | What it is |
|---|---|
arabic-autocorrect  |
Probabilistic autocorrect for Arabic text — edit distance over Arabic corpora |
NLP-Arabic-Datasets  |
Curated Arabic-language datasets for NLP tasks |
CBOW-For-Arabic-Language  |
Word2Vec (CBOW) for Arabic, built from scratch |
