We cut fueling fraud 82% with a camera and OCR, not a bigger model
Fleet-fuel fraud has a shape. Here's the shape, why after-the-fact detection kept losing, and what changed when verification moved to the moment before the wallet is charged.
Senior Software Engineer · Fintech & Payments · Cairo
Startups bring me in to build the part that can't fail.
Eight years across fintech, crypto and high-traffic platforms — usually owning the services everything else depends on. Right now that's the wallet behind Egypt's largest fleet-payment network: the ledger, the transaction path, and the fraud check in front of both.
Wallet platforms, double-entry ledgers and transaction workflows. Money systems have one hard requirement — the numbers reconcile, every time, including the times something breaks mid-flight.
Formance · Temporal · PostgreSQL
Event-driven architectures and the orchestration that holds a long-running process together across a dozen services. Service boundaries drawn so they still make sense a year later.
Kafka · gRPC · Kubernetes
Order matching for a crypto exchange, live market data, and high-traffic platforms where latency is the product and a bad design shows up immediately.
Go · Redis · ClickHouse
2018 – present
Six systems, one habit: the part that can't fail goes first.
Products I built or led the backend for. Each entry is the problem, what I built, and what changed.
I do my best work where a startup is outgrowing its first architecture.
Small teams, real traffic, and decisions that are expensive to reverse. I've spent most of my career in that phase — the point where a service that worked fine for a year starts costing money every time it breaks, and someone has to redesign it without stopping the business.
Much of it where correctness, latency and money meet: an edge case there isn't a bug report, it's a balance that doesn't add up. That means transaction consistency, idempotency, reliable event propagation, and workflows built to be resumed rather than restarted.
And because startup teams are small: service design, architecture review, and mentoring that leaves people able to maintain the thing after I've moved on.
Full history on LinkedIn
2026 – now
ExpertHub · GreenMobility, Copenhagen · remote
Financial data platform on Snowflake, telemetry-driven billing, ~40% lower cloud spend.
2024 – now
Octane, Cairo · hybrid
Own the wallet — ledger and transaction engine — behind Egypt's largest fleet-payment network.
2023 – 2024
Dinamo MEA · remote
Led a 5–10-engineer backend team: PalCards, Biddi Market, e-commerce platforms across MEA.
2022 – 2023
AtenTEC, Dahab
Led the backend team building Anchornize; on-prem → GCP migration.
2021 – 2022
Genwin, Ottawa · remote
Order matching and Kafka pipelines for a Solana creator-economy platform.
2018 – 2021
Zad Academy, Kuwait · remote
Full-stack for a global online academy.
Education B.E. Electrical, Electronics & Communications Engineering, Alexandria University (2019) · ALX / Holberton Software Engineering (2023) · Data Science Certificate, ExploreAI Academy (2024) · Programming with Google Go Specialization, UC Irvine (2023)
Highlighted where I go deepest
Notes from the part that can't fail.
Fleet-fuel fraud has a shape. Here's the shape, why after-the-fact detection kept losing, and what changed when verification moved to the moment before the wallet is charged.
Open to senior and staff engineering roles — remote, or with relocation — and to consulting with startups on payments infrastructure and distributed systems.
[email protected]The full history lives on LinkedIn.