AI-Native QA Engineer

Returning.AI; Singapore; On-site; SGD 70,000 - 110,000; Full time

Job summary

Returning.AI is seeking a QA engineer who tests with AI agents, not just manual clicks. You will own quality for our platform, building Playwright suites, API tests, evals, and reusable test playbooks that AI agents run around the clock.

You will work directly with the founder and engineering team in a small, high-ownership environment. This role emphasizes judgment, meticulousness, and the ability to guide AI-driven testing at scale.

Qualifications

• Real QA experience testing real products.
• Comfortable with Playwright and Selenium.
• Able to build test suites, not just execute cases.
• Able to measure quality with metrics and evidence.

Responsibilities

• Build the automated test suite with Playwright end-to-end tests and API tests.
• Write reusable playbooks and skills for agent-driven testing.
• Develop evals to catch drift in AI features and test coverage.
• Own release quality and triage bugs into reproducible steps.
• Raise the bar by creating scalable testing systems used by the whole team.

Skills

QA experience

Playwright

Selenium

AI testing

Meticulous

Clear writer

Education

Degree in CS or Information Systems

Tools

Codex

Claude

Cursor

Job description

The short version

We need a QA engineer who tests with AI agents, not just with their hands. You will own quality for our platform and build the system that keeps it high: Playwright suites, API tests, evals, and reusable test playbooks that AI agents run for you around the clock.

This is not a manual-testing role with some automation sprinkled on top. If your instinct when facing a 200-case regression pass is to open a spreadsheet and start clicking, this is not for you. If your instinct is to make an agent generate, run, and maintain those cases while you decide what actually needs testing, keep reading.

You do not need to be a heavyweight coder. AI agents write most of the code now. What we cannot get from an agent is your judgment: critical thinking, meticulousness, and an eye for detail. That is what we are hiring.

You will work directly with the founder and the engineering team. Small team, real clients, high ownership.

About Returning.AI

We build gamified loyalty and retention systems for CFD brokerages. Our clients are enterprise, paying $60-144K per year. We are a 10-15 person team based in Singapore, profitable, and growing.

Our product sits inside broker client portals and drives trader engagement through rewards, challenges, and lifecycle campaigns. Real integrations, real money, real commercial stakes. A bug a client finds before we do costs us trust in a niche market where word travels fast.

Requirements:

• Degree preferably in comp science or information systems


• Familiar with Codex / Claude / Cursor / other harnesses


• Built / Vibe-coded stuff before


• Meticulous

Why this role exists

Our engineers already ship with AI every day, and the product moves fast. Testing has to move at the same speed, and no human team can key through screens fast enough to keep up.

We want one person who owns quality and multiplies themselves with agents. Test coverage should grow every week without headcount growing with it. Today, too much of our testing is manual and too little of it is repeatable. You change that.

What you will actually do

Build the automated test suite. Playwright end-to-end tests for the flows that matter, API tests for the services behind them, all running in CI on every release. You decide what gets covered first, you keep the suite green, and you make sure a green run actually means something.

Write reusable playbooks and skills. Package testing knowledge so an agent can run \"smoke-test staging\", \"verify the reward redemption flow\", or \"regression-pass the campaign builder\" on demand or on a schedule. A playbook you write once should run a thousand times without you.

Build evals. Where the product uses AI, you measure it: build evaluation sets that catch when an AI feature's behavior drifts after a prompt or model change. And you eval your own testing agents too. You should know when your AI testers are missing bugs, and prove coverage with evidence instead of vibes.

Own release quality. Decide what blocks a release and what does not. Triage bugs into clean reproductions a developer can act on without a call. Report quality in numbers the team trusts.

Raise the bar for everyone. When you build a playbook, eval, or agent workflow that works, it becomes something the whole team uses. You are building a testing system, and the system should outlive any single test case.

What good looks like in six months

• Every release gets a full automated regression pass without anyone having to ask for it.


• Agents run scheduled test passes and find real bugs before clients do, including while you sleep.


• New features arrive with test cases generated the same day the spec lands.


• Bugs escaping to production are trending down, and you can show the numbers.


• The suite grows every week, and maintaining it takes you hours, not days, because agents do the repair work.

Who you are

Real QA experience. You have tested real products and know how bugs hide. You have built or run test suites, not just executed cases someone else wrote.

Comfortable with Playwright and Selenium. Comfortable, not expert. The agents write most of the test code. Your job is to direct them, read what they produce, spot a test that passes without proving anything, and unstick an agent when it loops. If you already lean on Cursor or Copilot to build things you could not code alone, that is exactly the profile.

Meticulous, with an eye for detail. You notice the button that moved, the date that is off by one, the totals that do not add up. An agent can generate a hundred test cases in a minute. You are the one who notices the important case that is missing.

AI-native in practice. You already use AI tools daily and can show it. If you have ever made an agent write, run, and fix tests for you, we want to see it. If you have not done that yet but you are the kind of person who automated half your own job, tell us about that instead.

Professionally skeptical. You know AI output can be confidently wrong. You verify. A green test that asserts nothing bothers you more than a red one, because the red one is at least telling the truth.

A clear writer. Bug reports, test plans, and coverage summaries that a non-technical founder can read and act on.

High agency. There is no QA manager above you handing out assignments. You see the risk, you cover it, you report what you did.

What this is NOT

• Not a manual testing role. Manual exploration is a tool you use, not the job.


• Not an ML or model-training role. You are using AI to test product, not building models.


• Not a test-case-management role. Nobody here wants a beautifully organized TestRail instance with no automation behind it.


• Not process-heavy. No test plans that take longer to write than the feature did to build.


• Not remote. Singapore-based. We need you in the timezone and occasionally in the room.

Strong signals

• You have introduced automation or AI testing into a team that did not have it before.


• You can show a real example of an agent-assisted testing workflow: the prompts, the output, what you kept, what you threw away.


• You have found bugs that mattered and can tell the story of how.


• You have opinions about flaky tests and what to do about them.

Weak signals

• Your automation experience is one Selenium course and a certificate.


• You say you are \"into AI\" but cannot show how it changes your actual testing work.


• You measure QA by test case count instead of by bugs caught and releases protected.


• You need a finished spec before you can start testing anything.


• You accept whatever the agent produces without checking it.

What you get

• Direct founder access. Your findings reach the decision-maker the same day.


• A green field. You are building our AI testing system from the ground up, your way.


• An AI-forward environment. We do not just allow AI tools, we expect them, and we pay for the tooling you need.


• Real stakes. Our clients have thousands to hundreds of thousands of traders. The quality bar you hold protects real revenue.


• Competitive compensation. [SALARY RANGE], depending on experience.