Muhammed Yıldız · QA Engineering · Remote
I've tested software since 2021, across e-commerce, CRM, payments, telecom and speech AI. A test run at the end catches what it can. A system decides what reaches the end at all: exploratory work where judgment matters, automation where repetition does, both on the same pipeline. I build that, and it keeps working once I'm gone.
Open to freelance and consulting work.
The record
testing software, since 2021
companies shipped for
test frameworks built from an empty repo
disciplines, manual and automation, run by one person
Services
Four things I get brought in for. Each one is something I've already done in production, not a package invented for this page.
Most teams don't need more tests. They need a setup where writing one is cheap and the result is trusted. At BoxCommerce I built that from scratch: Playwright and TypeScript for the interface, Postman for the API layer, page objects, fixtures and parallel runs, all wired into CI so every pull request answers for itself.
Payments, pricing, logistics, OAuth and calendar sync at BoxCommerce. R&D documentation and reporting at innoscripta. Telecom CRM at ETIYA. Speech and LLM products at SESTEK. In these areas a missed edge case doesn't stay a bug. It becomes a refund, a support queue, or a customer who doesn't come back. I concentrate coverage where failure is expensive instead of spreading it evenly.
At ETIYA I moved the team onto a hybrid way of working, exploratory and automated running together instead of a standing argument about which one counts, and set up the Jenkins pipelines that kept it running without me in the room. A suite starts decaying the moment nobody owns it, so I hand over the habit, not just the repo.
I've never treated these as rivals. Exploratory testing finds what nobody thought to specify; automation makes sure it stays found. Running both myself means what I discover on Tuesday is covered by Friday, with no handoff in between. AI sits in the toolkit on the same terms: sharp when you know exactly what you're asking it for, noise when you don't.
And the reason the advice lands
Through Prime Picks I design, build and release apps end to end: Android, web, PWA. It keeps me honest about the cost of every bug I file, and it changes what I choose to raise.
How I work
Testing is mostly judgment under time pressure. These are the models I fall back on when there isn't time to check everything.
Inversion
Anyone can confirm that software works. I start from the failure and work backwards to the path that causes it.
5 Whys
A bug report that stops at what happened gets fixed once. Push it to why it happened and it gets fixed for good.
Second-order thinking
The question is never only whether the patch works. It's what the patch moves downstream of itself.
Entropy
A suite nobody owns is already rotting. Keeping it trustworthy costs energy every sprint, and spending that energy is the job.
Asymmetry
Coverage spread evenly is coverage wasted. I weight it toward the handful of paths where failure is genuinely expensive.
Honesty
Green does not mean ready. If a release isn't safe, that's the report, however inconvenient the timing.
Adapted for testing from the mental models collected at mmpractices.com.
Work
innoscripta · Remote from Turkey
2026 · Present
Full-time contract QA on Clusterix, the R&D documentation and project tracking platform built by innoscripta in Munich, where companies record research hours, tasks and costs as the work happens so the result holds up to an audit. Manual and automated testing side by side, running alongside my BoxCommerce work.
BoxCommerce · Remote
2024 · Present
Brought in to build QA for the e-commerce and CRM side, and stayed to run it. I wrote the Playwright and TypeScript framework, added Postman API automation underneath it, and covered payments, pricing, logistics, OAuth and calendar sync. Exploratory testing and automation run on the same GitHub Actions pipeline.
ETIYA · Hybrid
2023
Ran manual testing for a telecom CRM and moved the team onto a hybrid model. Built the Selenium and TestNG regression suite, and set up the Jenkins pipelines that kept it running without me in the room.
SESTEK · Hybrid
2021 · 2023
Tested LLM speech products: speech to text, text to speech, and the analytics dashboards built on top of them, across both desktop and web applications. API automation with Postman, web automation with Selenium, working in Agile on Azure DevOps. This is where manual and automation stopped being two separate jobs for me.
Open source
Public on GitHub. If we work together, this is roughly what lands in your repo in the first week.
Writing & speaking
“Enhancing QA Work”, on getting more out of QA by putting human judgment and automation in the same loop instead of picking a side.
Studio & products
Prime Picks is my small app studio. Everything here I designed, built and shipped myself.
Let's talk
Send me the product and where quality hurts right now. If I'm not the right fit I'll say so, because that costs us both less than finding out in month two.
