I design and deploy enterprise AI — RAG assistants, Text-to-SQL platforms, agentic data analysis, computer vision and OCR pipelines — from architecture through deployment and monitoring.
I build enterprise AI systems across LLM engineering, RAG, agentic workflows, computer vision, OCR, forecasting and automation — with a focus on letting people query knowledge, databases, documents and spreadsheets in plain language.
I own projects end to end: architecture, data preparation, model and tool selection, retrieval design, backend and API development, testing, deployment, monitoring and delivery to business users.
Every system is built around data privacy, reliability, explainability and measurable business value.
Enterprise assistants connecting language models to internal documents, policies and departmental knowledge.
Workflows that classify intent, route tasks, query databases and explain results.
Real-time vision for security monitoring, recognition and attendance, plus OCR and document-intelligence pipelines.
Bilingual retrieval, summarization, classification and employee support.
FastAPI, Docker, Azure, local LLM serving, logging and monitoring.
A RAG assistant unifying policy search and assistant chat over a shared knowledge base, backed by OCI GenAI and Claude.
Converts natural-language business questions into governed SQL over Oracle sales data.
Fully offline bilingual RAG assistant for internal HR-policy retrieval.
Routes spreadsheet questions to SQL, visualization, report and data-quality workflows.
On-premise vision platform for fire/violence detection, plate recognition and attendance.
Whisper-based transcription, extraction and summarization pipeline.
Classifies Facebook and Instagram comments across brands with a fine-tuned BERT model into a 9-category taxonomy.
Designing and deploying enterprise AI across language models, data analysis, computer vision, OCR, forecasting and automation.
Bilingual (Arabic/English) OCR extracting batch, manufacturing-date, expiry-date and quantity fields from product labels.
Custom Prophet-style Ridge regression and LightGBM ensemble models for supply-chain demand forecasting, with LLM-generated root-cause explanations.
Ingests product spec documents and uses Claude to generate GSO-compliant bilingual marketing layouts, gated by an automated compliance check.
Power BI dashboards for inventory turnover, fulfillment and operational KPIs.
Gesture recognition driving an Arduino robotic arm.
AraBART fine-tuned on 10,000+ Arabic documents.
Speaker segmentation for noisy, multi-speaker Egyptian Arabic audio.
Real-time detection paired with a farmer-facing chatbot.
Object detection, OCR, face recognition and text-to-speech in one assistive app.
Deep-learning web app classifying cotton grade from imagery.
Enterprise RAG, agentic workflows, Text-to-SQL, computer vision, OCR and AI automation — for enterprise teams or as a freelance engagement.