The craft, up close

How the work actually looks.

“Most of my work is making release day boring. The good kind of boring.”

— how I think about the job
Live release sign-offs
100% held the gate · every cut
cypress/support/commands.ts
// one helper · every checkout spec calls it · no copy-paste
Cypress.Commands.add('seedAndSignIn', (role: UserRole = 'returning_customer') => {
  cy.task('db:reset');
  cy.task('db:seed', { role, fixtures: 'checkout' });
  cy.intercept('POST', '/api/auth/session').as('getSession');
  cy.session(role, () => {
    cy.apiLogin(role);
    cy.visit('/');
    cy.wait('@getSession');
  });
});
flake · before → after
15% <6%

Assertions split into classes, data-test selectors centralized, state seeded via DB tasks - not the UI.

Selected work

What I get hired to fix.

/01

A flaky suite nobody trusted, rebuilt to 700+ green tests.

A team had inherited a thin, flaky test suite everyone had learned to ignore. I rebuilt it end to end - page objects, dedicated assertion layers, centralized selectors, database-seeded state - and grew it to 700+ tests that run green every night in under two hours. Flake dropped about 60 percent, and the next hundred tests cost a fraction of the first. Their own engineers write tests now.

  • Cypress
  • Playwright
  • TypeScript
  • CI
  • self-hosted
/02

AI features shipping blind, given an evaluation layer that catches bad output before users do.

A product was shipping features on Claude and GPT with no way to know when quality slipped. I built the evaluation layer: output scored against ground truth, hallucination and prompt-injection checks, RAG grounding tests, all wired into CI so a bad prompt or a model upgrade fails the build instead of production. Built on the Anthropic and OpenAI APIs with Promptfoo and DeepEval.

  • Anthropic API
  • OpenAI API
  • Promptfoo
  • Red-teaming
  • RAG
/03

Quality owned on a live payments product, from zero process.

Brought in as the only QA on a payments product with no test process in place. I owned all of it - test strategy, manual and exploratory passes, end-to-end automation, release-readiness criteria, and the go/no-go gate. Card payments, Tap to Pay, refunds, 3DS, and transfers. Multiple critical bugs caught every release, before a single user hit them.

  • Payments
  • Manual + E2E
  • Release management
  • PCI-aware

Inbox open

Let's make your next
release boring.

Mail is the fastest way to reach me. Tell me what you are shipping and where it hurts, and I will tell you how I would test it.

Email tykhonkozachenko@gmail.com

Remote · worldwide · CEST +352 661 566 607

T.