About me
I am a product manager who came up through sport. Ten years of competitive volleyball taught me that talent is cheap and habits are expensive, and I run product the same way: short cycles, honest feedback, no hiding behind a deck.
These days I work on AI products, mostly on the unglamorous parts. Making a model useful inside a real workflow. Deciding what not to build. Sitting with customers until the actual problem shows up.
Outside of work I run AI workshops, drink an unreasonable amount of coffee, and answer basically every message I get. If you are reading this and wondering whether to reach out, reach out.
- 175% ARPU growth from a pricing model I redesigned
- 4 days from idea to a deployed product with MyMasareef
- 4 years of top flight volleyball, most of them as captain
- Cairo home base, working with teams anywhere
Side projects
Things I built on nights and weekends, because the fastest way to understand a tool is to ship with it.
MyMasareef
Live Product | mymasareef.com
An AI spending analyzer for CIB Egypt customers. Upload a statement, get your money explained back to you in plain language. Financial data never leaves the browser.
How it works: Pattern matching against 90+ known Egyptian merchants. Gemini Flash for unrecognized entries. User override with a crowdsourced merchant dictionary that improves over time. All financial processing stays in the browser.
Key decisions: Email gate comes after the user sees value, not before. Crowdsourced dictionary reduces cost while improving accuracy. Debit cards got a dedicated experience.
Tech: React, TypeScript, Tailwind, Supabase, Express, Gemini 2.5 Flash, Claude Code, Recharts, Framer Motion. Built in 4 days.
Call Transcript Analyzer
Internal Tool | GitHub
A Claude Code pipeline that turns hours of customer calls into themes a product team can act on. Built it because reading transcripts was eating my week.
Pipeline: Transcript ingestion from Clari Copilot, multi-pass extraction (features, bugs, complaints, praise), categorization by product area, cross-transcript deduplication, Confluence-ready report generation.
Key decisions: Used CLAUDE.md as a context engineering layer. Multi-stage processing beat single-pass for accuracy. Chose Confluence output because the team already lived there.
Tech: Claude Code, Claude API, Confluence. Built in 1 day.
El Shabab League
Live Product | elshabab.alyelazab.com
Mobile-first prediction game for a 2026 football knockout tournament. Players call the exact score, the goalscorers, and the 15-minute window each goal lands in. A Supabase Edge Function pulls live results and scores everyone automatically on a shared leaderboard.
Key decisions: Score first, scorer second, because the final score is the call that settles an argument. Automated scoring on a schedule instead of manual updates after every match. Real squads and countries, no licensed logos or crests.
Tech: React, TypeScript, Supabase, Edge Functions, PWA, Cloudflare Pages. Built in 1 day.