Pavlo Barchuk
QA Engineer  ·  France, remote  ·  Open to roles

Quality for software
that thinks.

I test AI products, and I build the agentic systems that test them, from a raw defect to deployed infrastructure. From running my own service business, to enterprise QA, to production AI quality. The proof is not a title. It is what I have built.

QA lead AI quality systems AI-driven automation Multi-agent pipelines
Pavlo Barchuk, QA engineer
The path

Operator, then builder.

Not a straight line of one job title. A move from running a business, to testing enterprise software, to keeping AI products honest in production, to building the systems that do the testing.

2016 to 2021

Operator

Founded and ran a service business end to end, with my own engineers. Sales, process, delivery.

2021 to 2025

Enterprise QA

Test strategy and regression on a Big Four client’s audit and financial-data platform.

2023 onward

AI quality

QA lead on a production AI product. Built an AI-driven test-management system from scratch.

Now

Builder

Several large systems in Claude architecture, with cross-model second opinions and structural gates.

Approach

Traditional QA breaks for AI.
I rebuild it.

The old playbook was written for deterministic systems. Non-determinism, hallucinations, prompt injection, evaluation gaps, and drift do not fit it. AI products do not need less QA. They need QA moved up a level: from running cases to designing the system that runs and checks itself. The automation, the frameworks, the CI: written by the systems I design and direct, then verified and shipped. The leverage is in the architecture, not the typing.

Done” from an AI is not “correct.” A 200 OK is not the work landing. The check itself has to be checked, by a second independent pass.

  • 01Multi-agent test pipelines. Specialist agents that handle non-deterministic output, not one prompt doing everything.
  • 02Evaluation frameworks. Precision, recall, hallucination rate, drift. Quality criteria for stochastic output.
  • 03Quality gates, prototype to production. The irreversible rules live in the structure, not the prompt. Described is not enforced.
  • 04Cross-model second opinion. One model checks another. A degraded run raises instead of pretending to be healthy.
  • 05Self-improving infrastructure. Corrections are captured and become rules, so the system does not drift back.
Skillset

What I work with.

Hands-on QA on the manual side, plus the automation, CI, and frameworks I design and direct through AI. The methods that hold quality together, and the AI-quality layer where I go deepest.

01

Testing

FunctionalExploratoryRegression User-story validationAPI checksCross-browserSmoke / sanity
02

Methods & practices

Test strategyQuality gatesDefect lifecycle Requirement analysisRisk analysisCI / CDQA reporting
03

AI quality & automation

LLM evaluationMulti-agent test pipelinesAI-driven UI & API automation Automation framework designHallucination & driftPrompt-injectionCross-model reviewSelf-improving infra
04

Stack

Claude / Claude Code Jira · Xray Playwright Java · SelenideRestAssured · Postman Python Salesforce GitHub PostgreSQL · SQL
Systems I built

The proof, not the title.

One typed architecture over Claude Code, built solo for a production product, then carried into other domains. Open any system to see what it does, where the value is, and how it is built. The instance is replaceable. The method is the asset.

Experience
  • Apr 2023
    present
    QA Engineer (contract)
    Production AI product · remote
    QA lead on a production AI product. Built an AI-driven QA system from scratch that stands up regression and API automation, designed the automation framework the team's Java suite runs on, defined release test strategy, ran quality gates and sign-offs, mentored the QA team.
  • Oct 2021
    Jun 2025
    QA Specialist
    Big Four client platform · audit & financial data
    Test case design, execution and review, defect reporting, root-cause and requirement analysis. Functional, exploratory, smoke, sanity, regression, UI, cross-browser, API testing.
  • Sep 2022
    Jun 2023
    QA Engineer, mobile (part-time)
    iOS / Android
    Requirement analysis, test case design, bug reporting, weekly QA reporting, release sign-off. Push-notification and back-office testing.
  • Sep 2016
    Oct 2021
    Founder
    Service business · Kyiv
    Ran all business processes end to end: customer support, service delivery, finance, strategy, sales and engineer-team training, marketing and partnerships. Operator before builder.
Education
Kyiv Electromechanical CollegeAutomated control systems
Hillel IT SchoolInformation technology
Languages
NativeUkrainian
WorkingEnglish (intermediate)
OtherFrench (beginner), Russian
Certifications

Practice first,
theory reading it back.

On the Anthropic Solutions Architect track. I built the real systems first, the certification confirmed the method.

Claude 101 Claude with the Anthropic API AI Fluency for Small Businesses Introduction to agent skills Anthropic Solutions Architect track
Pavlo Barchuk at his desk
About

The person behind the systems.

I find the thought through working, not through buzzwords. Technical knowledge commoditizes. A sense of where a product is going does not. The systems I build are less about typing the code and more about deciding what to build, why, and how to keep it honest under pressure.

I sit in the middle of the river, at anchor, and watch how it moves. Not drifting, not fighting the current. I move when I see where.

Contact

Open to QA roles where AI is the job.

If you are bringing AI into your product and want quality that holds in production, I would like to hear about it. The fastest way to reach me is below.