Projects I document as I work on them: Power BI, Python, automation, and analytics. These double as my portfolio.
Built to replace a daily 15 to 60 minute payment check for 4 coaches and 300+ students
A jiu-jitsu coach wanted an easier way to check payments. Underneath it was a bigger need: one place for four coaches to see everything about 300+ students.
Two series, 13 chapters, English and Arabic, read in 20 countries
IlmiTales is a free, interactive storybook in English and Arabic that teaches children how AI works. Building it tested whether I understood AI well enough to explain it to a child, and whether what AI built for me was actually right.
Seven source systems into one live dashboard, built end to end
After owning the business logic of an enterprise footfall model, I built an entire analytics platform myself: synthetic data, medallion pipeline, star schema, and a React dashboard. With AI as the pair-builder and my judgement as the safety net.
Offline AI for confidential documents: answers in English and Arabic, with the source behind every answer
I built a document assistant that runs on a laptop with no internet, answers in English and Arabic, and shows the exact line it used. Here is how, and the one problem that took the longest.
Galaxy schema over ~500K synthetic rows, ~45 DAX measures, 3 audience pages
A portfolio Power BI project on fully synthetic data. One model that answers both clinical and donor questions across a 31-country humanitarian health program.
Margin variance cut from 48% to under 2% after two years unsolved
A root cause analysis across a Medallion data platform. How I traced a long-standing cost gap to two hidden causes and worked with engineering to fix both.
2.5 days of work down to under 3 minutes. 109,406 KWD saved. CTO Award.
An in-house Excel, VBA and Power BI tool that automated 400+ data quality rules, replaced a six-figure platform, saved 109,406 KWD, and earned a CTO award.
Nine years of study habits, 2,563 sessions, one live dashboard
I gave 2,563 focus sessions from 2017 to 2026 to Claude and got back a live dashboard and a video. The data showed me something I hadn't noticed about how I learn.
One trusted footfall number across 70+ brands and 3 regions
Door sensors, AI cameras, restaurant POS guest counts and web sessions, across the Middle East, Turkey and Central Europe. I owned the business logic that made them one trusted number.
~30K row flat export rebuilt into a 4-table star schema
A hierarchical Odoo POS export, nearly 30,000 rows deep, rebuilt into a four-table star schema and two dashboards. Including the measure that catches a cashier whose errors cancel out.
