I test software for a living, break into systems for fun, and use AI like a craftsman uses a tool - deliberately, only when it's the right one, and honestly about when it isn't.
01 · About
I'm a QA Engineer at XOGO since Sep 2022 - manual and automated testing, cross-platform debugging, and making sure software does what it says before it reaches a customer.
TryHackMe Jr Penetration Tester path, Advent of Cyber 2025 (24 challenges). Breaking things properly makes me better at making sure they don't break in production.
Started with Perplexity in March 2023, before the hype cycle. Claude and Claude Code now, real shipped projects, no fake screenshots.
02 · Built under a clock
A GenAI-powered recovery companion and caregiver support app, built solo in one hackathon sprint - with a real e2e test suite, not just a demo.
03 · Credentials
04 · Shipped work
A local-first MCP server that bridges AI assistants to a real, logged-in Chrome session over Chrome DevTools Protocol - built because Playwright-driven browsers kept getting blocked by Cloudflare and other anti-bot layers. Any MCP-compatible AI can use it, not just mine.
Built with Perplexity in 2023, before I called myself an AI toolsmith - my first real project built alongside AI. Kept live as a marker of where this started.
A community project, not a solo personal tool - daily-ritual tracking (PWA + Android) built and maintained for a wider group of practitioners.
05 · Skills
06 · Services
Most AI usage is wasteful - too many tokens spent for too little output, prompts that fight the model instead of working with it, workflows that would take half the time with the right setup.
Solo builder, a team, or a whole business - doesn't matter. I audit how you're actually using AI - Claude or otherwise - and show you where the waste is: token spend, prompt structure, tool/agent design, and where AI shouldn't be used at all. The goal is less spend, more signal, and no gimmicks.
