What this talk is about
An agent will write you a test in seconds. Whether that test is worth keeping is a different question. Most teams end up with suites that pass locally, break in CI, and that nobody can map to actual coverage — so the tests get skipped, then deleted, and the automation effort resets to zero.
This session is about the part that starts after the agent hands you a test: the setup, the architecture and the practices that make AI-generated tests hold up in a real repository over time.
Testers and QA automation engineers first, but useful to full-stack and backend developers who own their own coverage. Pen testing and component-level unit tests are in scope too.
Meet the speaker
Yevhen has spent over a decade building test automation from the ground up. Most recently, as Chief Quality Officer on a high-load fintech and crypto trading platform, he owned the automation architecture across UI, API, microservices and blockchain components, and built the CI/CD quality gates and observability-driven validation around them.
What you’ll take away
An end-to-end AI automation setup
Playwright driven through MCP servers and skills — how the pieces actually fit together.
A prompt turned into a working test, live
Plus an honest read on how much time it saves versus writing it by hand.
Architecture that survives CI
How to organise skills, what a code graph is and why it matters, and where Obsidian fits.
Practices that hold up
Monorepo test layout, and what the good and bad patterns each actually buy you.
Previous sessions
19 June 2026
Context Engineering
3 July 2026
System Design
4 June 2026
