DEFINITION
Engineer who builds the production layer of AI products — turning model capabilities into reliable, observable, actually useful systems.
Led a team of 4 engineers building automated project management for multiple ministries. Built alazharmemory.eg — a digital archive for the Al-Azhar library, launched at the Cairo ICT international fair.
Led a team of 3 on Vodafone's Remote SIM Provisioning platform — eSIM management for 20,000+ users worldwide. Early eSIM days: building the spec while building the product.
Built data pipelines connecting media content into graph representations — giving journalists tools to surface connections across stories they wouldn't have seen otherwise. Cloud analytics for content performance at BBC scale.
Refactored the Order Transmission Engine to Grade A and brought data quality incidents to zero. Built cloud-native data integration pipeline with 100% consistency between core streams.
Rider-driver matching engine — real-time location, supply-demand balancing across Gulf traffic. Partnered with data scientists to improve dispatch efficiency.
Lead AI engineer across the AI org. AIDA multi-agent insurance assistant (symptoms → provider booking). Nexes-ADE document extraction at scale (OCR + LLMs). LangGraph-powered Quote Compare engine. Designed GitOps foundation with ArgoCD, Helm, AWS EKS + Azure AKS, and full LGTM 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.
Cognitive aptitude test prep platform
Practice app for the Criteria Cognitive Aptitude Test (CCAT). Timed test simulations, detailed answer explanations, and performance analytics. FastAPI backend with Next.js frontend, full auth, admin panel, and progress tracking across sessions.
Daily AI-generated morning command center
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.
Fixed-scope AI and engineering sprints
Four fixed-scope sprints: AI workflow automation, data & ops dashboards, landing page + MVP build, and infra/cloud rescue. Fixed price, short delivery windows, you own the output. Email to start — no portal, no retainer.
Data-to-dashboard in 48 hours
Send a spreadsheet, get an executive dashboard plus three action items in 48 hours. Done-for-you, human-powered audit for SMB operators. Fixed-price tiers from $500 (one dataset) to $5,000 (pipeline + decision layer). Three slots a month.
No-upload CV bullet review — fake-door experiment
A focused 6-filter pass on 2–3 of your CV bullets, so each line survives the recruiter skim. No upload, no form, no account, no payment — you email me the bullets you choose to share. Testing whether there's demand for a human bullet critique before building anything.
Natural language product search
"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.
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.
3D multiplayer social environment
Real-time player synchronization, spatial audio, chat, and a persistent world. Built on PlayCanvas engine with Node.js backend. Currently in security audit and stability phase.
Private baby cry analyzer for parents and caregivers
iOS app that listens to a baby's cry, classifies what they need (hunger, discomfort, fatigue), and gives actionable care advice. Private inference — audio is analyzed on a private server and never stored. Expo/React Native, Supabase backend. Free on the App Store.
Zero-config AI code review GitHub App
Install once, get inline comments on every PR — catches bugs, security issues, and style problems. Free tier: 5 PRs/month. Pro: unlimited @ $8/month. Launching Q3 2026.
AI wardrobe stylist app
Photo-based wardrobe cataloging with AI outfit suggestions from your own closet. CLIP embeddings for visual similarity matching, garment lifecycle management with pg_cron. Expo + Supabase with pgvector for semantic search. Phase 1 in progress.
CONSULTING & CONTRACTS
I work with companies building AI products that need to actually ship. Multi-agent systems, document extraction pipelines, production ML infrastructure — the kind of work where the demo needs to become a reliable system people trust.
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.