18 results in numbers, from ten years of work.

Leadership asked two questions: why is development so slow, and why can't we onboard new operators? The audit found the answer: launching a brand was a manual process, and team processes wasted time. After automating brand setup, a new brand went live in one day instead of a week.

A large client had been complaining for three months and the team couldn't find why. I went through 288K error events myself, found the causes in two days, and the team shipped four fixes.

Bonus wins could reach real-money balances. I ran the fix as one project: 22 days from start to production, with the gap closed for good.

The bonus team shipped with too many errors and unstable features. After I reworked how it delivered, releases went out every week, rollbacks happened no more than once in two months, and urgent production fixes came down to about one per five features.

Each new game-provider integration took about a month and two people. After rebuilding the process it took about two weeks and one engineer.

Critical outages kept coming back and root-cause reports ran up to 17 times over their deadline. I traced each outage down to the code, ended each with a structural fix and a mandatory pre-production check, and set up a live tracker for every open report. Incidents went from about five a week to one, and that one no longer critical.

Engineers spent hours on routine configuration tickets. I put an AI coding agent into the workflow: it prepares each change as a ready merge request, and engineers only review it. Time spent on these tickets fell by about 30%.

When an outsourced engineer wasn't working out, I replaced them quickly instead of waiting and hoping. Utilization of the contracted team went up by 70%, so the company paid for work actually done.

Operators in Asia waited seconds for every response from a platform hosted in Europe. I moved 9 services into a second region in Singapore, and response times there dropped to milliseconds.

As the platform's FinOps lead I found about $55K of savings across 12 initiatives and wrote the budgeting process for the next year's infrastructure.

Clients could only reach the back office through a VPN. I led the programme that removed it safely: penetration-test fixes, vulnerability clean-up, an identity-system upgrade and secrets moved to a vault.

Launching payments in a new regulated market took about two months. I rebuilt the payments platform so each market runs its own production setup, and a new market went live in about three weeks. Along the way the team grew from 2 to 20 engineers, with 50+ payment providers, and a cost check in planning stopped 25% of unprofitable features before development started, saving about six person-months a year.

The platform couldn't keep up with the data coming from trucks, and nobody was fully sure the numbers were right. We moved it to an event-driven design with separate read and write paths, and it went from 10,000 to 30,000 events per second.

The ad-auction service topped out at 10,000 requests per second and the business needed 15,000. After a redesign it handled 17,000. Separately, new tools for campaign managers took over ad setup and data sync between platforms, removing around 40% of their manual work.

About a third of problem bookings were getting lost, because nobody could see the whole process. My team of seven rebuilt it as a state machine on Camunda, so a case can no longer drop out unnoticed. In six months losses fell to about 5%.

When COVID stopped face-to-face sales, the client had only an internal system and no way to sell insurance packages. I audited that system, then built an online sales platform that sends every application straight into it, growing the team from just me to 10 engineers. Live in about six months; around 35% of new applications moved to the new channel.

The course platform couldn't run live streams or keep recordings, so the academy couldn't sell live formats. We added both, and the academy started selling a new type of product.

A floral studio I co-founded. I built its whole operating system with AI agents and no dev team: an ordering bot, an online shop, stock and every order in one place, automatic replies to reviews.