Evidence before theory
No fix goes near production until the cause is proven. A plausible explanation that nobody tested is how the same incident happens twice.
About
I lead development and maintenance for merchants whose platforms carry real money every hour of the day. The interesting part is rarely the feature. It is the integration that retries, the patch nobody read or the fault that only shows up at peak.
I started with PHP 4, on shared hosting, in an era when deployment meant FTP and hope. Everything since has been a slow lesson in what actually keeps systems alive: boring releases, honest logging, and knowing which failure modes you have chosen to accept.
Today I am Technical Lead at Deploy Ecommerce, working across Adobe Commerce, Magento and Laravel. My week splits between keeping stores trading, connecting them to the ERP, warehouse and third-party systems that run the business behind the storefront, and building internal tools that take manual process off people who have better things to do.
The work I get called for most is diagnosis. When a fault crosses system boundaries, when two vendors each insist the problem is at the other end, when it cannot be reproduced on staging, that is usually when the phone rings. I work the same way every time: gather evidence before forming a theory, rank the competing explanations, then prove one of them wrong.
I also write and publish open source, contribute to Magento core and answer more questions than I ask in the Adobe Commerce, Mage-OS and wider PHP communities. Most of what I release started life as something a merchant needed on a Tuesday.
How I work
Three rules, learned the hard wayNo fix goes near production until the cause is proven. A plausible explanation that nobody tested is how the same incident happens twice.
Integrations fail in the middle. Assume duplicates, assume replays and make recovery an ordinary Tuesday operation rather than an escalation.
Pipelines and agents remove the typing. They do not remove accountability for what ships. Somebody named has to hold the rollback decision before the window opens, and it should not be whoever is watching the graphs.
Away from the keyboard
Confirm or replaceLong days and high ground. The best architecture decisions I have made were made about four hours into a walk.
A practice in consistency, which is the same skill as shipping every week rather than heroically once a quarter.
Endurance, engineering and escape, in roughly that order depending on the weather.
Racing
PlannedResults
Placeholder rowsActivities
Strava profile →Sample activities. Pulled from Strava when the site builds, so the list is as fresh as the last deploy. Powered by Strava.
Active in the Adobe Commerce, Mage-OS and wider PHP communities: packages, patches, talks and answers.
Architecture reviews, integration second opinions and talks. Contact.