$ whoami
Full Stack AI Engineer β Egypt (EET)
$ focus
RAG systems Β· LLM apps Β· Arabic OCR/NLP Β· FastAPI backends
$ currently_building
LMS backend Β· Egyptian ID OCR Β· Hybrid RAG engine Β· Quranic KB| Area | What it means in practice |
|---|---|
| π RAG Systems | Hybrid retrieval (dense + BM25) Β· rerank Β· eval Β· observability |
| πͺͺ Arabic OCR | Detection β OCR ensemble β postβprocessing β structured JSON APIs |
| βοΈ Backend | FastAPI Β· clean architecture Β· Docker Β· CI/CD Β· auth/JWT |
| π Multilingual AI | Arabicβfirst UX Β· embeddings Β· custom normalization & tokenization |
Goal: Turn noisy ID images into clean, validated JSON.
How: YOLO detection β crops β PaddleOCR + EasyOCR ensemble β rules & confidence gating.
Result: ~92% fieldβlevel accuracy on real samples, ready for bank/fintech KYC APIs.
Tech: YOLO Β· PaddleOCR Β· EasyOCR Β· OpenCV Β· FastAPI
Goal: Make it easy to ship grounded Q&A over mixed Arabic/English corpora.
How: Dense + BM25 hybrid search, reranking, smart chunking, source citations, eval scripts.
Result: <2s endβtoβend latency on typical corpora with noticeably fewer hallucinations.
Tech: Python Β· Qdrant Β· BM25 Β· LLM APIs Β· LangChainβstyle pipelines
Goal: Backendβfirst LMS that can grow into a full product.
How: Modular FastAPI services, JWT auth, roleβbased access, media storage on Azure Blob.
Result: Ready to plug into any web/mobile frontend and scale with real users.
Tech: FastAPI Β· MongoDB Β· Azure Blob Β· Docker Β· JWT
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
β π LMS Backend FastAPI Β· MongoDB Β· Azure Blob Β· JWT β
β πͺͺ Egyptian ID OCR YOLO Β· PaddleOCR Β· EasyOCR Β· Validation β
β π Hybrid RAG Qdrant Β· BM25 Β· Rerank Β· Eval pipeline β
β π Quranic KB RAG Β· Arabic NLP Β· Knowledge graphs β
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ



