Portrait of Valentin Lutchanka

Hands on Agentic QA Engineering Manager
Verisk Underwriting Solutions · Málaga, Spain

ValentinLutchanka

AI quality engineering, measured in hours given back to the team.

I've put AI agents to work on the repetitive parts of testing: refining requirements, triaging defects, and most of what happens in between. So far they've taken 3,500+ hours of manual effort off my team. We spend that time on risk-based and exploratory testing, which is still where people find the bugs that matter.

3,500+ hours
Labor hours I saved by Agentic automation across QA activities and operations. Measured against the manual baseline it replaced, Verisk, 2023 — present.
Manual baseline With agents

What I changed at Verisk

Each number is measured against the manual baseline it replaced.

  1. 3,500+

    working hours saved by introducing the automation with agentic approach and autonomous agents across QA activities and operations

    Baseline: manual effort · Method: DMAIC

  2. 12

    SDET and UAT QA analysts led in API, UI and performance testing

    Acceptance and UAT environments · Agile

  3. +70%

    testing efficiency increase from an in-house API test cases generator, 700+ hours saved yearly

    Two projects · 700+ hours a year

  4. 2×

    faster deployments through automated CI/CD quality gates

    Bamboo · smoke and regression suites

Where agents sit in my QA team's processes

I led a team to implement autonomous agents across our QA processes. Agents with project and product knowledge take the high-volume work while people make the calls that need judgement.

  1. PnPlan

    Requirements analysis & refinement

    AI-assisted requirements analysis implemented with human in the loop, grooming and refinement. Ambiguities, gaps and testability risks come up during refinement, while they are still cheap to fix.

  2. AAssemble

    Test design & automation

    Agent-assisted test authoring on a unified Playwright with K6 framework, plus test code generators that produce API tests and assertions with full response coverage and contract validation.

  3. IInspect

    Test results analysis

    AI-powered analysis of test runs separates real regressions from noise, so the team acts on signal and the feedback loop stays short.

  4. StSelect

    Defect classification & triage

    AI-powered defect classification and triaging routes issues to the right owner with consistent severity, cutting manual triage effort.

  5. TLTransport loaded

    QA operations & reporting

    I led research and development of AI agents to take over recurring QA operations, with unified real-time reporting that notifies stakeholdersright in time.

The codes are therbligs @Frank and Lillian Gilbreth.

How I run AI adoption

I started in QA in 2018 as a manual tester. That's where I learned where testing time really goes, and it's why I moved into automation, then SDET work, then leading a team. Now I manage how we use AI in testing. Having done each of those jobs, I know which parts are repetitive enough to give to an agent. Triage, sorting through test results and the first read of new requirements go to agents, while the decisions stay with people. I measure a baseline before I introduce anything, because if the gain can't show up as a measurable outcome, I don't trust it.

  1. DDefinePick the repetitive work an agent can take
  2. MMeasureBaseline the manual effort
  3. AAnalyzeFind where the testing time really goes
  4. IImproveIntroduce the agent
  5. CControlControl the result

Measure first

I am Lean Six Sigma Green Belt certified. Each agent developed in my team follows DMAIC: baseline the manual effort, introduce the agent, then control the result.

People in the loop

I follow the approach where agents absorb the repetitive work. Analysts move up to exploratory, risk-based testing and to reviewing what the agents produce.

Enable the team

I led 10+ internal trainings a year on current automation and AI practices, and hands-on mentoring for analysts moving into agentic automation. Four team members certified in Playwright and Langchain.

One stack, one standard

I consolidated four insurance and underwriting projects onto a single automated framework I managed my team to build from scratch. That removed tools spread and made quality comparable across projects.

Career

  1. 2023 — Present Verisk Underwriting Solutions

    Agentic QA Engineering Manager

    • Lead a team of 12 SDET and UAT QA analysts across API, UI and performance automated AI powered testing for Acceptance and UAT environments in an Agile setup.
    • Introduced the agentic approach and autonomous agents backed with LangChain across STLC stages: AI-assisted requirements analysis and refinement, AI-powered defect classification and triage, and test results analysis. 700+ QA and Operations hours saved.
    • Managed a team to build an Agentic test automation framework from scratch (Playwright, K6) and integrated four insurance and underwriting projects into it.
    • Managed development of various AI API test code and data generators: +70% testing efficiency on two projects and 3,500+ labor hours saved yearly, freeing time for exploratory and risk-based testing.
    • Published various npm test-automation support packages powering Bamboo CI quality gates with smoke and regression suites. Deployment time cut by 2×.
    • Established unified, real-time test reporting and stakeholder notifications.
  2. 2022 EPAM · Málaga

    QA Automation Engineer

    • Migrated UI and API regression suites from WebdriverIO and Mocha/Postman to Cypress, unifying the project's stack.
    • Risk-based test plan improvements shortened test execution by 20%.
    • Docker, GitHub Actions and Jenkins integration gave a consistent test environment and cut maintenance time by 20%.
  3. 2018 — 2022 Virona IT · Remote

    Manual / Automation QA Analyst

    • Functional and non-functional testing of sports-domain products across the full testing life cycle, using WebdriverIO, Postman/Newman and Mocha/Supertest.
  4. 2009 — 2018 Club de Hockey · Benalmádena

    Manager

    • Nine years of business operations: customer relationships, sales and service, conflict resolution and office management.

Education

  • Bachelor of Business Administration Dutch Delta University, Netherlands · 2004 — 2006
  • DevOps with Docker, Full Stack CI/CD & Containers University of Helsinki · 2023
  • Test Automation program EPAM University · 2022 — 2023

Stack

  • AI & agents

    MCP (Playwright, LangChain, Chrome DevTools) · LangGraph

    Agents across 5 STLC stages

  • Test automation

    Playwright · K6 · NewmanAlso Cypress, WebdriverIO, JMeter, Mocha, Supertest, Jest / Vitest, Pytest

    Team-wide framework built from scratch, 4 projects on it

  • CI/CD & cloud

    Jenkins · Bamboo · GitHub Actions · DockerAlso GitLab CI, AWS

    Automated quality gates, 2× faster deployments

  • Programming

    JavaScript / TypeScript · Python

    Agentic development in both languages

  • Method

    Lean Six Sigma (DMAIC) · ISTQBAlso risk-based testing, Agile / Scrum

    Every agent rollout baselined first

Certifications

Show all certifications

Building an AI quality engineering function?

I am open to Head of AI Quality and QA engineering leadership roles. The fastest way to reach me is email or LinkedIn.