I work at the point where AI governance stops being a policy document and becomes something you can test. Founder and CEO of Life From Water, an Egyptian NGO working on water access and climate resilience; Ashoka Fellow; MBA. I also build the internal systems those organisations actually run on.
What I spend my time on
- Governance you can execute. Turning what an auditor asks for into permission checks, audit trails, and negative-case tests — the kind that fail loudly when a safeguard isn't really enforced.
- AI in operations, with a person in the loop. Systems where a model proposes and a named human confirms, permissions are re-checked at the moment of execution, and the consequential write is attributed to whoever approved it.
- Bilingual (EN/AR) products. Arabic and RTL as a first-class requirement, not a translation pass.
- Evidence and evaluation. Two randomised evaluations with J-PAL MENA; teaching AI governance and responsible AI to 27,000+ learners.
Things you can open right now
| AI Governance Roadmap OS | Free bilingual course — 8 stops, 39 lessons, four decision tools. NIST AI RMF, ISO 42001, EU AI Act, OWASP. |
| lifefromwater.org | Life From Water — bilingual donation platform |
| Impact Anchor | AI and evidence advisory for mission-driven organisations |
| Mohandes AI | Practical AI for Arabic-speaking engineers |
| mindkata-gate0 | Governance you can execute, literally: a scope lock a build checks. 61 lines of CI that fail the build when an agent drifts outside approved scope — with the ADRs, threat model and data classification behind it. Try the prototype. |
| motion-study | Video-based motion and time study for assembly and fabrication lines. Every comparable tool is commercial; this one is Apache-2.0. Live demo report. |
On the empty-looking contribution graph
Most of what I build lives in private repositories — internal systems for NGOs and clients. That work isn't public and shouldn't be. What's here is the small public slice.
I build with coding agents, and I own the problem definition, the architecture, the threat model and the verification. If something I've written is worth discussing, I can explain every design decision in it without the agent in the room.
Reach me via LinkedIn or lifefromwater.org.