Remote Ft- Senior Quality Engineer (AI Product) - Salary Upto 2000 USD

Monthly Gross

● Reports to: Chief AI Officer (CAIO), Japan HQ

● Works with: the product owner who runs UAT today, the AI engineering lead, and the consultants who onboard hotels

The Quality Engineer owns "green means true" for AI products. You decide what gets tested, how and how fast across our repositories; you run the UAT rounds and turn every finding into a traced, filed issue; you write or drive the fix with AI coding agents; and you report status in a form the CAIO can act on in two minutes.

You are an engineer first, not a tester. You read TypeScript and Python, you work through coding agents every day, and you judge what their green checkmarks actually prove. In our codebase the defects that mattered were found by booting the real application, driving it over real HTTP as a staff user and as a hotel user, in Japanese and English, and scanning live data — while the automated suites stayed green. That loop is yours to own, systematise and speed Up. We are hiring judgment and leverage, not clicks. Your value is measured by escaped defects, cycle time from finding to fix, and how much of the verification work you turn into harnesses and runbooks that agents and colleagues can run without you. That makes written communication a core technical skill: findings traced to a file and line, status a busy reader finishes in two minutes, pull requests that carry their own evidence.

What you'll do

1. Own the test strategy per repository. Decide what is a unit test, what is end-to-end, what needs a live-HTTP check and what needs a data scan. Remove tests that prove nothing (empty test bodies, flaky suites) rather than letting them read as coverage.

2. Build and run the end-to-end harness. Playwright (or equivalent) across our two web services: staff role and hotel role, Japanese and English, desktop and 390px mobile. Runs in CI on every pull request.

3. Run UAT rounds end to end. Plan, fixtures, walkthrough script, a findings document with expected and actual behaviour, every finding traced to a file and line before it is filed, issues filed the same day with severity and a Definition of Done.

4. Write the fix for what you find. Inside the team's delegation rules, ship the fix pull request with a regression test that pins the defect. When a finding is a scope or product question, file a decision ticket instead of guessing.

5. Test with AI coding agents, daily. Brief agents to generate and extend tests, review their output as carefully as a senior engineer reviews a junior's, and keep the prompts and checklists that make them reliable. We use Claude Code today and stay tool-agnostic: Codex, Cursor, Gemini CLI, Copilot and whatever ships next are fair game.

6. Guard data quality. Run and extend the fleet data-quality scan over stored hotel fact sets and warehouse drops: dry-run first, per-hotel backups, guarded writes, verify the live page after any repair.

7. Be the adversarial reviewer on agent-written pull requests. Reproduce the claimed verification, check the honesty rules on any hotel-facing copy, check Japanese and English parity, and check that no real customer or cloud identifier reached a tracked file.

8. Report clearly. One plain-text weekly status (fixed, being built, queued, awaiting decision, with issue numbers) and verification evidence pasted into every pull request you touch.

9. Turn QA documents into runbooks. Checklists, test plans and runbooks that a new agent session or colleague can execute without you in the room.

10. Measure the bar. Escaped defects versus defects found in UAT, flaky test.

Yêu cầu công việc

What you bring

• 3+ years in software engineering, SDET or QA automation, with ownership of a test strategy — not only execution of someone else's test cases.

• Hands-on experience with AI coding agents in real work: you have generated, reviewed and shipped tests or fixes this way. Any tool counts (Claude Code, Codex,Cursor, Gemini CLI, Copilot). Be ready to walk us through a repository or test suite you built this way.

• Reads and writes TypeScript/Node and Python well enough to write a fix pull request with tests. Playwright or an equivalent browser-automation tool. HTTP and REST debugging. Git, GitHub pull-request workflow, GitHub Actions.

• Test-design judgment: risk-based coverage, exploratory testing, reproducing defects over the real interface, data-quality checks. You can explain in one paragraph why a green suite is not proof, and you review agent output as carefully as a senior engineer reviews a junior's.

• Cloud basics on Google Cloud (Cloud Run, Cloud Storage, BigQuery) or an equivalent, with credential and identifier discipline: secrets and customer identifiers never enter a tracked file.

• Strong communication, in writing first. Findings with expected and actual behaviour, traced to code; a weekly status a busy reader finishes in two minutes; review feedback people act on rather than resent.

• A bias for shipping: small team, fast cycles, clear written updates.

How to apply. Send a CV and one written answer: describe a defect that a passing test suite missed, how you found it, and how you proved the fix. The process has four steps and we decide within two weeks of application:

1. CV and written answer review.

2. Take-home, time-boxed to about three hours, on a seeded build we provide: drive the app over HTTP in Japanese and English, find three real defects, write a findings note traced to file and line, and open one fix pull request with a regression test. Any AI tool is allowed; submit your prompts and transcripts with the work.

3. A 60-minute pairing session with the CAIO, live in Claude Code on a real open issue.

4. A conversation with the product owner about how UAT runs today, then an offer.

Quyền lợi

• Thưởng: Paid based on KPI Achievements