Selected work

Curated highlights from tooling and automation work. Structure: problem, approach, tools, and outcome.

Virtual Kanban boards in Software Development

Problem: Physical Kanban boards may not be sufficient for modern mobile work where teams need virtual, always-available workflow visualization.
Approach: The thesis reviews virtual Kanban board availability and examines three products against Kanban rule implementation, usability, cost, and two case studies for different team structures and needs.
Tools: Comparative analysis, case studies, Kanban methodology
Outcome: All evaluated virtual solutions were easy to use and provided added value over physical boards, but feature alignment with Kanban principles varied and tool choice remained context-dependent.
Link: Open project

Problem: Failing Robot Framework runs can be noisy and time-consuming to triage, especially when failures span logs, traces, and flaky behavior.
Approach: Executes the Robot suite, parses failed tests from output.xml, enriches analysis with newest Playwright trace data, applies secret redaction, and requests structured AI analysis in a typed schema.
Tools: Python 3.10+, Robot Framework, Playwright trace extraction, OpenAI structured outputs, JSON artifacts
Outcome: Produces machine-readable reports with classification, confidence, root cause, retry decision, and fix suggestions; supports CI with bounded cost controls via max-failures and message limits.
Link: Open project

Problem: Mobile UI automation is often fragmented across scripts, inspectors, and manual steps, creating slow and brittle feedback cycles.
Approach: Implemented a TypeScript VS Code extension that registers 7 Copilot-callable Appium tools and talks to Android devices through raw W3C WebDriver HTTP (no Appium SDK dependency).
Tools: TypeScript, VS Code Language Model Tools API, Appium, W3C WebDriver HTTP, UiAutomator2, artifact capture (.appium-artifacts)
Outcome: Turns natural-language prompts into executable Android test actions in VS Code and produces screenshots/XML traces that improve triage speed and reproducibility.
Link: Open project