m
mahmoudrfaey

Mahmoud Refaey

@mahmoudrfaey

AI Engineer building RAG systems, LLM agents and Arabic NLP

Egipto
Árabe, Inglés
Parte de la información aparece en idioma inglés.
Sobre mí
I build LLM apps that work on real, messy data: RAG chatbots over your documents, LangGraph agents, and Text-to-SQL tools that let your team query a database in plain language. Much of my work is in Arabic. At GBG I helped take an Arabic banking RAG chatbot from 79% to 97% faithfulness, mostly by fixing OCR on scanned PDFs and improving retrieval. Stack: Python, FastAPI, LangGraph, Azure OpenAI, Qdrant, PostgreSQL, Docker, LoRA fine-tuning. Tell me what you're building and I'll tell you how I'd approach it.... Lee más

Habilidades

m
mahmoudrfaey
Mahmoud Refaey
desconectado • 

Revisa mis servicios

Desarrollo de chatbots de IA
I will build a rag chatbot for your documents with scanned PDF support in any language
DeUSD 75 / proyectoMás información
Consultoría tecnológica con IA
I will evaluate and fix your rag chatbot hallucinations and retrieval accuracy
DeUSD 95 / proyectoMás información

Porfolio

Experiencia laboral

Brands

AI Engineer

Brands

Jul 2026 - Present • 4 mos

Working on an Arabic RAG chatbot over a regional bank's internal PDFs. I worked on most of the pipeline: EasyOCR ingestion, cleanup, chunking, hybrid retrieval (BM25 plus Azure OpenAI embeddings in Azure AI Search, merged with reciprocal rank fusion), query rewriting, cross-encoder reranking and evaluation. Better OCR on the Arabic PDFs and stronger retrieval helped raise faithfulness from 79% to 97% and correctness from 61% to 85% (measured with DeepEval). I also co-built Chinhook, an open-source Text-to-SQL agent sponsored by GBG. I designed the LangGraph workflow (query rewriting, schema retrieval from Qdrant, SQL generation, and a verification step before any answer goes out), restructured it so each user connects their own database and model without stored credentials, and deployed the internal version on Azure with Docker.

Beetleware

AI Engineer Intern

Beetleware

Aug 2025 - Jan 2026 • 5 mos

Worked remotely on NexaOS, an AI-powered web desktop app, on both the AI side and the full-stack side. I built an assistant agent for the marketing team and a RAG chatbot that answers questions about the system's content and documents. I also added AI features inside NexaOS apps, like AI text generation in the Notepad-style editor. Along the way I fine-tuned models, built data pipelines and contributed to open-source projects, shipping through CI/CD.