QA Engineer building AI-powered test automation — 13+ years across telco, banking & insurance, now turning that domain depth into autonomous testing tools.
I test billing platforms, CRMs, and self-service portals for a living — and build agents that test (and heal) them for fun. Currently exploring how far LLMs can go in closing the loop between bug found → bug reproduced → test fixed, without a human re-typing a selector ever again.
13+ yrs QA → Python / Playwright / pytest → AI-assisted automation → ???
| Project | What it does |
|---|---|
| 🔥 PhoenixQA | Self-healing test framework — LLM diagnoses broken selectors (Shadow DOM, dynamic attrs) and repairs them, Safe or Autonomous mode, learns from every decision |
| 🐛 defect-pilot | AI agent that reproduces a Jira bug and generates a Playwright retest script from the ticket alone |
| 🧪 llm-qa-toolkit | LLM-as-judge framework for testing chatbots in regulated industries — hallucination, prompt injection, regression |
| 🏗️ qa-automation-framework | The skeleton underneath it all — Playwright + pytest, POM/SOM, SQLAlchemy, enterprise telco/CRM/billing context |
These aren't isolated demos — they're one ecosystem. qa-automation-framework is the chassis, PhoenixQA keeps its selectors alive, defect-pilot closes bugs against it, llm-qa-toolkit applies the same AI-QA thinking to a different problem: testing LLMs themselves.
- I only put a technology on my CV once I've actually built something with it — no checkbox-collecting
- Domain knowledge is the differentiator I lean on: I've debugged billing migrations and credit-risk logic that most automation engineers have never seen
- AI is a tool I use daily (Claude, Cursor) rather than a buzzword on a slide. Most projects above include an AI-assisted component that can be inspected directly in the code.
- Pragmatic over theoretical: I'd rather ship a working POC with a
LEARNINGS.mdthan a perfect architecture diagram that never runs
- Self-healing automation strategies — the core problem PhoenixQA is solving
- LLM evaluation and benchmarking — Safe vs Autonomous healing accuracy
- Agentic QA workflows — agents that read a bug ticket and act on it (defect-pilot)
- Local vs cloud LLM trade-offs for regulated/enterprise test data
Python Playwright pytest SQLAlchemy Jira API Ollama Claude / Anthropic API GitHub Actions Allure
nofluffjobs profile · ⚽ Fan of AI, Python, and Spanish football