Account-management banking app, anonymised. Illustrative curve.
Selected work
Automation, performance and AI, built as systems rather than scripts
Four engagements that shaped how I work, plus a set of public, runnable demos you can clone and execute in minutes.
Open-source demos
Small, self-contained repositories: local demo apps, synthetic data, ready for CI. Each one shows a pattern I use in real projects.
- View repository (opens in a new tab)Playwright web E2E demoAutomation architecture
Login, cart and checkout journeys against a local demo app. HTML reports, failure diagnostics, GitHub Actions.
- Playwright
- TypeScript
- GitHub Actions
- View repository (opens in a new tab)Selenium framework demoAutomation architecture
Java + Selenium 4 framework skeleton: Page Objects, driver factory, screenshot artifacts, embedded demo server.
- Java 21
- Selenium 4
- JUnit 5
- Maven
- View repository (opens in a new tab)API testing demoPerformance engineering
Health checks, JSON contract validation and p95 latency baselines against a synthetic HTTP service.
- Java 21
- JUnit 5
- HttpClient
- Jackson
- View repository (opens in a new tab)JMeter performance packPerformance engineering
Reusable JMeter plan, CSV scenario data and a Python mock API with controllable latency for safe baseline load tests.
- JMeter
- Python
- CSV scenarios
Proof
Outcomes, not adjectives
What the work actually changed, plus a CV you can read in two minutes.
Minutes
instead of hours of performance analysis per test cycle
AI reporting pipeline
Android + iOS
one framework, both platforms
Banking app, hybrid WebView
0
manual result syncs to Zephyr
Every CI run publishes itself
2 tiers
of AI: commercial models for code, local models for sensitive data
GitHub Copilot + Ollama
Curriculum vitae
Two pages, print-ready. Bank Pekao S.A. (since 2019), T-Mobile, Tsunami, Agora. Over 15 years across QA, automation and web.
How I work
Quality as systems engineering
I treat test automation as an architecture problem, not a scripting task. Frameworks that outlive the next redesign, performance data that reaches decision-makers, and CI signals that people actually trust. That is the job.
Assess
Understand the system, identify risk areas, and define what quality means before building automation.
Architect
Design framework patterns that remain stable across product change, not scripts that break on the next release.
Automate
Focus on high-value flows and integration points where automation improves delivery confidence.
Report
Turn execution data and performance results into structured signals that support engineering decisions.
Applied AI
AI in quality engineering: the right model for the data
Day to day at the bank I work with GitHub Copilot (Chat, agent mode and the coding agent) and Microsoft 365 Copilot, on Claude, GPT and Gemini models. Where test data is sensitive, analysis runs on-premises through Ollama. Either way, AI speeds up engineering work and does not replace judgment.
$ qa-agent analyze results/run-142.jtl --baseline results/run-138.jtl
> parsed 48 210 samples | 6 transactions | 0.42% errors
> p95 latency: 812 ms -> 1 034 ms (+27%) on POST /transfer
> throughput stable: 118 rps vs 121 rps baseline
> error cluster: 91% HTTP 502 from gateway between 14:32 and 14:35
> likely cause: upstream timeout during deploy window (see release notes)
OK report written -> reports/run-142-summary.docx (local model, no data egress)
I treat AI as an engineering tool. It is useful when it improves signal quality, reduces manual effort, and keeps decision-making grounded in data.
Test result analysis
Parsing logs, clustering failures, and surfacing root causes faster than manual triage.
- Summarising test failures across suites
- Grouping similar errors by pattern
- Identifying likely root causes from stack traces
Performance report interpretation
Turning JMeter results and performance baselines into actionable observations.
- Explaining throughput and latency anomalies
- Comparing runs against historical baselines
- Generating stakeholder-readable summaries
Agentic engineering with Copilot
GitHub Copilot as a daily engineering partner: Chat, agent mode and the coding agent.
- Writing and refactoring test code in agent mode
- Delegating scoped tasks to the coding agent, reviewed as pull requests
- Custom instructions, prompts and agents for the QA team
Choosing where data goes
Commercial models for code and documentation, local models when test data is sensitive.
- Claude, GPT and Gemini models through GitHub Copilot
- Microsoft 365 Copilot for documentation and reports
- Ollama on-premises for sensitive test results
Stack
Tools I work with daily
Core technologies and methods across automation, performance, delivery tooling and applied AI.
- Java
- Selenium
- Appium
- Bitbucket
- Apache JMeter
- Python
- TypeScript
- Jenkins
- GitHub Actions
- Maven
- JUnit 5
- Jira
- GitHub Copilot
- Claude
- Gemini
- Ollama
- Docker
- Git
- Android
- iOS
- pandas
Automation engineering
- Java
- Selenium WebDriver
- Appium
- Playwright
- Maven
- TestNG / JUnit
- Page Object Model
- Data-driven testing
Performance engineering
- Apache JMeter
- CSV / JTL analysis
- Performance reporting
- Trend comparison
- Result visualization
CI/CD and tooling
- Jenkins
- GitHub Actions
- Build pipelines
- Jira integration
- Zephyr Scale
- Git workflow
AI and agent systems
- GitHub Copilot: Chat, agent mode, coding agent
- Microsoft 365 Copilot
- Claude, GPT, Gemini
- Ollama / local AI
- Agent orchestration
- Prompt and system design
Next step
Let's solve a quality problem
I help engineering teams ship with confidence: automation architecture, performance pipelines and AI-assisted quality systems. If your release cycle needs unblocking, let's talk scope.
Copy e-mail address
Łódź, Poland (remote)