DEFINITION
Engineer who builds the production layer of AI products — turning model capabilities into reliable, observable, actually useful systems.
Led team of 4. Automated project management for ministries. Built alazharmemory.eg digital archive.
eSIM platform for 20K+ users worldwide.
Data pipelines, graph representations for media, cloud analytics at scale.
Order Transmission Engine, data pipelines, zero incidents.
Rider-driver matching, real-time location systems.
Multi-agent systems (AIDA), document extraction (Nexes-ADE), LangGraph, full observability stack.
I studied biomedical engineering at Helwan University in Cairo. People sometimes call that a "non-traditional path" into software, but honestly, it's the best thing that happened to my career. When you've modeled biological systems, you develop an instinct for complexity, feedback loops, and things that fail gracefully — which is basically what good software is.
My journey started at the Egyptian MCIT, then wound through Vodafone, BBC, Talabat, and Careem before landing at Shory, where I now build AI-powered products for insurance in Abu Dhabi. Along the way, the stack evolved — JavaScript, TypeScript, Python — but the obsession stayed the same: making systems that are reliable, observable, and actually useful to real people.
I'm the kind of engineer who'll compare programming paradigms to Mozart vs. Beethoven in a team chat, share five AI papers before lunch, and then spend the afternoon making sure the monitoring dashboards are right. I believe a 99.9% success rate tells you more than 100% ever could.
/extract call. The unglamorous reliability work that makes throughput real.Self-improving healthcare AI agent platform
Multi-agent diagnostic consultation where a Patient, Doctor, Examiner, and Medical Director collaborate on cases. Dynamic specialty assignment — the case determines which 3 specialists participate. Evolutionary learning engine (SEAL) builds a case base from successes and validated failure lessons, with RAG retrieval for continuous improvement.
AI document extraction platform
Define document profiles with field schemas, upload files, get structured JSON back. Multi-engine OCR (PaddleOCR, Claude Vision, GPT Vision, Gemini) paired with LLM extraction pipelines. Full admin UI with audit logs, user auth, and real-time extraction stats.
"I need a modern blue sofa under $900" — GPT parses that into structured MongoDB queries, returns real products. Natural language search that actually understands what you want, not what you type. React + Tailwind frontend, Express + TypeScript backend.
AI-powered learning companion that turns any topic into structured knowledge. Conversational interface backed by a cognitive architecture that actually remembers what you've learned.
Morning command center — AI-generated city briefs across 4 cities and 8 news categories, live weather from 12 global cities, 20 daily interview Q&A across full-stack, ML, system design, and algorithms, plus a Hacker News AI digest. Auto-generated at 10 AM Dubai time.
Oil & gas production analytics dashboard — interactive charts, regional distribution maps, well location mapping with Leaflet, date/region/well filters, and a built-in chatbot for data guidance. Full-stack with FastAPI backend and Angular 17 frontend.
TikTok, but for your brain. A mobile app that replaces doom-scrolling with swipeable cards from great books and thinkers. Every card has an engagement gate — you can't mindlessly scroll. Daily cap enforces sessions, a visible knowledge graph grows as you learn, and every card traces back to its real source. React Native + Expo, Fastify API, Prisma.
ML-powered baby cry classifier — custom ONNX model with YAMNet embeddings that tells you why the baby is crying. Self-improving: low-confidence predictions get LLM second opinions, building a training dataset for continuous retraining.
Zero-config GitHub App for AI-powered code reviews. Install once, get inline comments on every PR — catches bugs, security issues, and style problems. Free tier: 5 PRs/month. Pro: unlimited @ $8/month.
Whether it's an interesting AI problem, a collaboration idea, or just a good conversation about why Beethoven's approach to composition mirrors functional programming — I'd love to hear from you. I'm also happy to share AI papers and articles; fair warning, I send a lot of links.