Services
What you can bring me.
Three services, one standard. Whatever the engagement, you get findings traced to their mechanism, and a report that says plainly what each claim rests on.
01 Manual and exploratory QA
Find what the test plan did not think of.
Good for: a release that feels risky, a product that has never had a tester, a redesign about to ship on phones, or a “works for me” that keeps coming back.
Exploratory testing is structured curiosity. I agree a charter with you (what we are looking at, what would worry us, what is out of bounds), then work through the product the way real people and real data will, across screen widths, light and dark themes, and awkward data shapes rather than the happy path.
What comes back is not a list of symptoms. Each finding is reproduced, rated by what it would cost you, and traced as far as the evidence allows toward the mechanism that produced it. A phone-width overflow and a dark-mode contrast failure are not two bugs; they are often one layout decision seen twice.
What you receive
- A written test charter before the work starts, so scope is explicit.
- Bug reports with reproduction steps, environment, severity, and the suspected mechanism, each marked with how sure I am and why.
- A short findings report: what was covered, what was not, the patterns behind the individual bugs, and what to fix first.
- Where it helps, a layout sweep across 4 screen widths × light and dark, the same discipline I run on my own releases.
How it runs
- Testing happens against your staging or test environment, or an isolated one I set up for the purpose. Your data never leaves the boundary we agree.
- Nothing from your product goes to a cloud AI unless it is anonymised first or you have said yes in writing.
- Short feedback loops: urgent findings reach you the day I find them, not at the end.
02 Test automation with Playwright
Checks that run on every change, and say what they prove.
Good for: a team shipping often with a thin safety net, a suite that is flaky and distrusted, or a product that must look right at every width and in both themes.
I build Playwright suites in TypeScript that your team can read and maintain: browser workflow checks for the journeys that matter, and layout checks that sweep widths, themes and data windows so visual regressions are caught by a machine rather than a customer.
Flakiness is a mechanism too. A test that fails one run in ten is telling you something about timing, state or environment. I would rather find out what than add a retry.
What you receive
- A Playwright suite in your repository, wired into your CI (GitHub Actions or similar), with a readable README.
- Workflow checks for the journeys we agree matter most, and a layout sweep across widths and themes where the product warrants it.
- A clear statement of coverage: what a green run proves, and what it does not. A passing suite is evidence for the cases it tests, not a promise about everything else.
- Handover notes so the suite outlives the engagement.
How it runs
- We start from your riskiest journeys, not from a coverage target.
- Suites run against the environments you choose. The browsers and the test runner live in your pipeline, not in a service of mine.
- Exploring and mapping the product comes before writing a line of test code. With your permission, I map it locally with LumenPort, my own testing workbench, before any checks are written, and your data stays local.
03 Automation and AI setup, local-first
AI that helps your team without taking your data with it.
Good for: a team that wants faster triage, reporting or test generation, but cannot (or should not) send code and customer data to a cloud model.
Local-first means the model runs on hardware you control, and anything sensitive stays there. Cloud AI has a place, for what is safe to share, but it should be a deliberate choice, not the default route for everything.
I run my own work this way: a large open-weight model on a home AI server answers the chat assistant inside my testing workbench, so client pages never go to a cloud AI. Every change to my software gets an independent review by a different AI model before it ships. I can set up the same kind of arrangement for you, sized to what you actually need.
What you receive
- A working local or hybrid AI setup for the tasks we pick: triage, report drafting, test-idea generation, log reading, or review of changes.
- Workflow automation (scripted, or with tools such as n8n) that moves findings from where they appear to where people act on them.
- A written policy for what may and may not leave your boundary, and the technical switches that enforce it.
- A review loop, if you want one: a second model independently checks the first, and fixes go round until the reviewer says ship.
How it runs
- We start with the data question: what is sensitive, where it may live, who may see it. The architecture follows from the answers.
- Everything is documented so your team can run and change it without me.
How I work The honest-evidence way
Four words that are never allowed to blur.
Every report, every status update, every case study on this site keeps these apart. When you read “done”, you will know which one it means.
- written
- The code or document exists. Nothing more is claimed.
- tested
- Automated or recorded checks exercised it, and the run is on file.
- deployed
- It is live, and the deployment itself was verified.
- independently checked
- Someone, or a second model, who did not write it reviewed it.
Fit
Before you write.
How big an engagement are we talking about?
Either shape works: a defined piece of work (a charter, a suite for the journeys that matter, a setup with a handover) or an ongoing arrangement. Say which you have in mind in the proposal form and we will shape it together.
Engagements start with a small fixed first step; I take on a limited number of clients at a time.
Will you sign an NDA? Where does my data go?
Client material stays on machines I control or inside your environment, analysed by local models. Anything that would go to a cloud AI is anonymised first, or you have said yes in writing. If your contract or NDA says no cloud AI, then no cloud AI.
Happy to sign a reasonable NDA. Contracts fall under the laws of Indiana, United States.
What do you need from me to start?
Access to a test or staging environment, a person who can answer product questions, and a sense of what would worry you most if it broke. The proposal form asks for exactly that.
What is not a fit?
Testing systems I do not have permission to test. Sending your users’ data to a cloud model because it is convenient. Reports that say “all green” without saying what was covered.