I write the schema, ship the code, and configure the server it runs on.
Scroll
About
I would rather own a system than a ticket. That has mostly meant drawing the boundaries myself — what gets built and what gets bought, what a client needs as against what they asked for — and then living with the consequences long enough to learn something from them.
Full stack engineer · BAT, Dubai (remote)
Full stack engineer · Lexzur — legal SaaS, Dubai (remote)
Senior full stack engineer · The iBoost — digital studio, Lebanon
Team lead & senior engineer · The Web Addicts — web agency, Lebanon
Freelance web engineer · Self-employed, Lebanon
LanguagesArabic · English · French · Japanese (N5 in progress)
Selected work
Context
Patchi has spent fifty years making chocolate a luxury gift rather than a confection, and sells everywhere from Beirut to Harrods. On a brand carrying that much heritage, the pages that sell nothing are doing the heaviest lifting: they have to hold a standard already set by the packaging, the boutiques and fifty years of gifting.
Contribution
I worked on the front end as one of the team, mostly on the informational side rather than the commerce flows — the brand, heritage and product storytelling pages, across six markets in three regions. The work was less about building features than about getting type, spacing and motion to a standard a luxury retailer would put its name to.
Note
The site has been through several iterations since 2022 and is not what the team shipped, though parts of that front end are still standing. This was shared work, and the credit is shared with it.
Nicotine brands operate under advertising and access restrictions that vary by market. The binding constraint is not who browses the catalogue, it is who registers: an underage sign-up is a compliance failure, and a self-declared date of birth is a field anyone can lie to.
Approach
The site runs on the platform BAT standardises on across its markets, with the capabilities the platform does not cover built as separate applications alongside it rather than bent into the theme. Identity checks go to a specialist provider. Customer data belongs to BAT's own system of record, and nothing reaches it until it is complete.
Decisions
Age verification went to a dedicated provider instead of a date-of-birth field, because a self-declared birthday does not survive a compliance review. That buys a third-party dependency in the middle of the signup funnel, and the drop-off that comes with it.
Outcome
Live since 2025, serving the catalogue, offers and promotions to verified consumers.
Houmata set out to sell furniture and home goods from many independent shops under one roof — a marketplace rather than a store. That is a different problem from the outside in: you are not building one catalogue, you are building a system that lets other people bring theirs, each with its own identity, pricing and stock.
Architecture
Two systems rather than one. The public marketplace is the half people see; the half that decides whether it works is the platform behind it, where a shop onboards itself — brand identity, product catalogue, pricing — with no developer in the loop. A marketplace lives or dies on how cheaply it can add the next vendor, and if onboarding costs engineering time then growth is capped by the size of the engineering team. Building that second system is what turned a website into a marketplace.
Delivery
I built it alone, from the initial data model through to launch, while leading the team on other projects at the same time. The client met daily — business logic, pricing formulas, technical feedback — so the specification was still moving while the thing was being built. Most of the difficulty was not the code. It was holding a fixed launch date against requirements that changed in the morning meeting.
Note
Houmata has changed considerably since: it now runs AI-assisted room planning and a different product model. What I built in 2023 is the marketplace it launched on.
A live storefront taking money in a regulated market. Once a store is trading the work changes shape: there is no greenfield, every change lands underneath customers who are mid-purchase, and a defect in the wrong place stops people buying rather than merely annoying them.
Approach
One platform pattern shared across BAT's markets, extended per market with custom applications rather than by bending the theme past its limits. On the platform team the job is less about building something new than keeping a commercial system healthy while it keeps trading — which means diagnosing what is actually failing rather than what appears to be.
Decisions
A custom API in the login flow was firing on every keystroke, tripping the platform's rate limit and locking customers out of their accounts. The obvious fix was to ask for a higher limit. The right one was to accept that the limit was doing its job and the caller was misbehaving: debouncing the input and windowing the requests removed the load instead of moving it somewhere else. Authentication is the worst place on a store to have an outage, because a customer who cannot log in does not come back later.
Outcome
The store trades reliably and defects are short-lived. Customers move through promotions, discounts and checkout without meeting the platform underneath.
Artists And More stages entertainment across the region — festivals, galas, product launches, mall activations — from a roster running to thousands of performers and an in-house production team. The site has to work as a shop window for that scale: a catalogue of talent, a gallery heavy with media, and a route from browsing to an enquiry, without any of it feeling like a directory.
Ownership
This was handed to me whole, as the only developer on it — and it was the second project of my career. Everything from the data model to the front end to the deployment was mine, which meant learning where the decisions actually were rather than being handed them. There is no faster way to find out which parts of a build you do not yet understand.
Note
The site has since been rebuilt, as sites are. What I shipped in 2021 ran the business for several years before that, which is the part I would point to.
The store had a customer base to grow, and the people worth reaching were existing smokers rather than newcomers to nicotine. That needs a reason to engage, not simply a storefront. The complication is that anything worth offering them sits behind the same wall that protects the brand: you cannot run a prize promotion for someone who is neither a registered customer nor verified as an adult.
Approach
A prize promotion built as a multi-step questionnaire, delivered as a custom application alongside the store rather than bolted into the theme. Entry requires a registered, age-verified customer, which puts the compliance gate directly inside the acquisition funnel — so how that interruption is handled becomes the design problem rather than a detail at the edge of it.
Decisions
The gate could not move, so the cost of it did. Someone who starts the questionnaire without an account is sent to register and then returned to the point they left, instead of being dropped back at the homepage to begin again. A campaign that loses people at the exact moment it asks them to sign up is not an acquisition campaign, and that one behaviour is the difference between a completed entry and an abandoned one.
Outcome
Live in the UAE, running as one of the campaigns the team ships against the same goal — giving existing smokers a reason to register, and a reason to come back.
