AI QA Engineer — Agentic Test Automation (Mid-level)

Threadcode Technologies · threadcodetech.com
Location: Kathmandu, Nepal (hybrid) · Type: Full-time · Openings: 1 · Reports into: QA / Engineering Enablement

About Threadcode & the role

Threadcode is a Nepal-based software engineering firm serving international clients. Our primary product, Event Logic, is a Swedish SaaS platform for corporate event management — white-labeled, multi-tenant, running across the Nordics and Europe in six languages.

We don’t just use AI to write code — we use it to test code. Our QA pipeline is agentic end-to-end: AI agents explore the application, draft Gherkin scenarios, implement Playwright step definitions, heal failing tests, and verify suspected regressions — while a human QA engineer designs the test strategy, approves what the agents produce, and attests to what a machine cannot judge. Every push is gated by an executed test playbook with recorded video evidence; every bug fix ships with a paired red→green proof that the defect existed and is gone.

This role is that human. You will not spend your days hand-writing test cases one at a time. You will orchestrate, review, and harden an AI-driven QA system — and be accountable for what it lets through.

If our QA Engineer posting is “prove it works,” this one is “run the system that proves it works.”

What you’ll actually do

  • Operate the agentic test cycle. Drive our LLM-agent pipeline (built on Claude Code) through its full loop: exploration → scenario drafting → step implementation → test run → heal → AI verify → green run. You are the approval gate at every stage — agents propose, you decide.
  • Shape test strategy, not just test scripts. Help author the QA playbooks (trigger surfaces, acceptance criteria, replication recipes) that agents and developers execute. You test the plan the AI didn’t think of.
  • Review AI-authored Gherkin and Playwright code against our conventions: declarative scenarios, no magic numbers, strict selector priorities (data-testid registries, i18n-safe role selectors, no fallback chains). Reject what doesn’t meet the bar — the agents learn from your corrections.
  • Run and improve the self-healing loop. Diagnose failing scenarios (selector drift vs. timing vs. real regression vs. app bug), supervise agent heal attempts, and enforce the discipline that distinguishes a flaky test from a shipped defect — including our “one heal at a time” rule that catches a single root cause masquerading as twelve failures.
  • Verify with evidence, not vibes. Execute pre-push validation gates: per-row pass/fail on every test recipe with video recordings, attached to Azure DevOps work items as reviewable testing evidence. Provide human attestation on the rows a machine can’t judge (visual quality, email content, business sense).
  • Regression-first bug handling. Every bug you file carries a deterministic replication recipe; every fix is proven by re-running the same recipe red→green. You’ll enforce fails-first regression tests so no bug silently returns.
  • Feed the learnings loop. When a test fails for a new reason, when an agent makes the same mistake twice, when a selector convention gap bites — you capture it as a structured learning and turn it into a durable rule, hook, or lint so the system never needs the same correction twice.
  • Test the hard stuff. Multi-tenant isolation, white-label theming across six client brands, a client × language i18n matrix, QAT/UAT dual environments, and AI product features themselves (LLM-powered agents inside Event Logic that need their own evaluation and guardrail testing).

Must-haves

  • 2–4 years of QA experience on web products, with real hands-on automation work (writing and maintaining tests, not just executing suites someone else built).
  • Genuine hands-on experience using LLM tools in your daily work — Claude Code, Cursor, Copilot, ChatGPT, or equivalent — beyond autocomplete: you’ve prompted a model to generate tests, reviewed its output critically, and can articulate where it fails.
  • Solid Playwright + TypeScript/JavaScript. You can read, correct, and extend page-object models and step definitions — you’ll be reviewing AI-written test code daily.
  • BDD/Gherkin fluency and an opinion on what makes a scenario good (declarative over imperative, behavior over choreography).
  • Test-design depth: pre-conditions, observable assertions, boundary analysis, and the judgment to know which requirement maps to a unit test, an integration test, or an E2E flow.
  • Professional skepticism toward AI output. The core skill of this role is not trusting green: knowing when a passing suite proves nothing and when an agent’s confident answer is wrong.
  • API testing (REST, auth headers, status-code discipline) and comfort reading network traces, console logs, and application logs to root-cause a failure yourself.
  • Structured written English — your bug reports, playbooks, and learnings are read by developers, agents, and stakeholders across three time zones.

You’ll stand out with

  • Experience with playwright-bdd or Cucumber-family toolchains.
  • Prompt engineering for test generation, LLM-as-judge / AI-verifier patterns, or evaluation of LLM-powered product features (guardrails, hallucination checks, tool-call correctness).
  • Exposure to agent frameworks or MCP (Model Context Protocol), CI/CD test gates, or self-healing test tooling.
  • Multi-tenant SaaS, localization/white-label testing, Azure DevOps test management.
  • Contributions to test infrastructure: selector registries, flake-mode catalogs, custom reporters, evidence tooling.
  • ISTQB is welcome but optional — a portfolio of automation you’ve built and AI workflows you’ve tamed counts for more.

What we offer

  • A QA role that doesn’t exist in most companies yet. You’ll work at the frontier of AI-augmented quality engineering, on a system already running in production — not a slideware initiative.
  • Real authority: you are the approval gate. Nothing ships past QA without your sign-off, human or machine.
  • Mature European product, modern infrastructure, and a team that treats QA as an engineering discipline, not a checklist.
  • Competitive Nepal-market salary, hybrid work, and afternoon Europe-timezone overlap leaving mornings for deep work.
  • Direct mentorship in agentic engineering practices you won’t find in a course — you’ll grow into this discipline on the job.

How to apply

Email your CV to info@threadcodetech.com with the subject line “AI QA Engineer”, and include at least one of:

  • An automation suite or test framework you built (repo link preferred);
  • A written example of an AI-assisted testing workflow you’ve used — what you prompted, what the model got wrong, and how you caught it;
  • Your best bug report — one with a reproducible recipe and evidence.

Applications are reviewed on a rolling basis.