Machine Learning Engineer at Certified IT Consultants CIC (2025-02 – Present)
- Built a Retrieval-Augmented Generation (RAG) and Context-Aware Generation (CAG) system using LangChain, ChromaDB, Neo4j, and Gemini API. Developed an end-to-end pipeline including document loading, chunking, embedding generation, semantic retrieval, and LLM-based answer generation.
- Constructed a Knowledge Graph by extracting entities and relationships from documents and storing them in Neo4j. Enhanced generation with context-aware retrieval by combining vector similarity search (ChromaDB) with graph traversal, enriching retrieved text with structured graph context to improve answer accuracy and reduce hallucinations.
- Developed a multi-path AI agent for a real estate chatbot that dynamically routes user queries to specialized processing pipelines based on intent. Implemented a database retrieval module capable of fetching property information through natural language queries, enabling users to search listings conversationally.
- Built a Retrieval-Augmented Generation (RAG) pipeline to answer questions using internal company policy documents and integrated an LLM-based conversational module to handle real estate inquiries with refuse an out of scope questions and improve the overall user interaction experience.
- Built an AI-powered reporting agent that interprets natural language queries and generates board-level analytical reports from structured company data. Implemented an intelligent query-to-database mapping mechanism that dynamically identifies the relevant database columns required to answer complex business questions.
- Developed a data aggregation and analysis pipeline that consolidates information from multiple fields to produce unified datasets suitable for executive reporting and decision support.
- Enhanced an existing OCR system for Egyptian National ID cards by extending support to process both the front and back sides of the ID. Implemented expiration date extraction to improve data completeness.
- Additionally integrated image similarity and visual verification techniques to detect whether an uploaded image corresponds to a valid Egyptian ID, or a non-ID image, improving the reliability of the identity verification pipeline.
- Developed a Qwen-14B based AI agent designed for meeting summarization and sentiment analysis. The system processes meeting transcripts to generate concise summaries while highlighting the overall discussion sentiment, enabling stakeholders to quickly understand key outcomes and insights without reviewing full conversations.
AI developer intern at Nahdet Misr publishing group (2024-04 – 2025-02)
Developed an Arabic NLP question generation system using transformer-based models.
Prepared
Arabic NLP datasets by converting raw paragraphs into sentence-level training data and applying preprocessing techniques. Implemented semantic similarity methods using Word2Word and Word2Vec to identify related words and contextual relationships.
Applied Named Entity
Recognition (NER) to detect entity types and improve contextual understanding. Finally, fine-tuned an AraBERT model to generate coherent and contextually relevant questions from input sentences.
AI developer intern at shadi systems company (2022-09 – 2023-07)
Designed and developed an OCR-based system for extracting structured data from the front side of Egyptian National ID cards. Built a computer vision preprocessing pipeline that removes background, applies filtering, and enhances text sharpness, combined with edge detection to isolate the ID card region, improving OCR accuracy. Implemented an OCR module to reliably extract key identity fields such as name, national ID number, and address.
The system transforms raw ID images into structured verification data suitable for automated identity processing workflows, ensuring high accuracy and robustness in real-world scenarios.
