AI-Native Quality Engineer (Zero-Touch Release, Security First)

MUST Company; United States; Remote; USD 90,000 - 130,000; Full time

Job summary

MUST Company is hiring a fully remote AI-native Quality Engineer to enable zero-touch releases with a security-first mindset. You will design automated gates, an oracle reference, and AI-assisted test flows to catch defects before release.

You’ll map existing checks, reduce human interventions, and measure lead time and cost per run across CI pipelines to continuously improve quality.

Qualifications

• 1–4 years hands-on QA, test engineering, security testing or software engineering on live products.
• You believe that better AI models increase quality and QA is career leverage.
• Security is your default: check input handling, access, secrets and hostile input.
• You have built an automated check that decides pass/fail with no human in the loop.
• You look for existing tools before building, and can say what you found and what gap remains.
• Daily, hands-on use of AI tools in your work, including corrections you made to AI results.
• Clear written English: reports and handoffs that engineers can act on without a call.
• You report numbers honestly and never claim a result you did not run.
• Bonus: experience with differential testing, image/file formats, mobile apps, Web3.

Responsibilities

• Build automated checks that run in one shot and catch what vibe coding misses, security first.
• Build a slow, exact reference (oracle) for shipped items and judge fast versions against it.
• Look for existing tools before building; fill only the gap.
• Measure human interventions, cost per run, lead time, and defects caught before release.
• Explain what checks cannot see and bridge that gap.
• Use AI daily to write checks, read logs, analyze results and generate test cases.

Skills

AI-assisted QA

Security testing

Automated quality gates

Differential testing

Property-based testing

Root cause analysis

Python

TypeScript

Test automation

Written English reporting

Job description

AI-Native Quality Engineer (Zero-Touch Release, Security First)

MUST Company · Bangladesh, Angola, Cambodia, Botswana, Chile, Ethiopia, India, Indonesia, Pakistan, Kenya, Nigeria, Vietnam, Philippines, South Africa, Nepal, Ukraine, Poland, Romania, Bulgaria, Moldova, Serbia, Georgia, Armenia, Albania, North Macedonia, Bosnia and Herzegovina · REMOTE

About MUST Company

MUST Company is a South Korea–based company building AI-native products. This role is fully remote and full-time, open to candidates living in any of the countries listed on this posting.

Why this role exists

As AI models get better, the value of code goes toward zero. What goes up is quality.

Quality always mattered, but only a few people on big projects could see it. Now everyone can. Vibe coding does not produce quality, and people feel it first through security problems. So security is the baseline, and the people who survive are the ones who move very fast and still raise quality very high.

We are hiring the person who makes that possible: someone who turns quality into the thing that lets us ship fast with zero, or close to zero, human intervention.

PhotoCraft is an open-source, Rust-based Photoshop alternative. It was built with clean-room engineering, not reverse engineering: from Adobe's public file-format documentation and observed behaviour, with no Adobe code, and with AI doing most of the first build.

Look at how its quality is proven, because this is the standard we want:

• A slow, exact reference renderer judges the fast one. Both render real Photoshop files in CI, and any difference is flagged down to the individual colour channel.
• Files must survive being opened and saved again. In its public write-up, 169 of 170 test files make the round trip, and the one that does not is visible, not hidden.
• Passing tests were not the finish line. Alpha testers still found 214 keyboard-shortcut failures that the unit tests never saw. Good QA knows what its checks cannot see.

Our reading of its public history is that commits land in one shot with no human intervention, and that each new app in the suite takes less time to build than the one before. It proves three things: a lower human gap, a lower cost, and a faster lead time. The core of all three is quality.

What you will do

• Make quality the reason we can ship fast with zero or near-zero human intervention
• Build automated checks that run in one shot and catch what vibe coding misses, security first
• Build a slow, exact reference ("oracle") for the things we ship, and judge the fast version against it, every run
• Look for what already exists before you build anything. Use it, and build only the gap
• Measure every claim: human interventions, cost per run, lead time, and defects caught before release
• Look beyond your own checks. Say plainly what they cannot see, and close that gap

What success looks like

• First 30 days: you have mapped what already exists, found at least one real quality or security problem in a live product, and shown it with a check anyone can run
• First 60 days: that problem is fixed and proven with before-and-after numbers, and a check runs on every release to catch it if it returns

  How we measure you

  Not by how many test cases you wrote or how many tools you built. By three numbers: human interventions per release, cost per run, and lead time, and by how many real defects were caught before a user ever saw them. Security issues that reach users count against you.

  How you will work with AI

  You use AI every day to write checks, read logs, analyse results and generate test cases. You stay responsible for the judgement: you check what AI gives you, you count how often you had to step in, and you can show where you corrected it.

  This is not the role for you if

  • You want to write test cases by hand and call that quality
  • You build a new tool before checking whether a good one already exists
  • You treat security as an extra step at the end

  Hiring process

  • Profile and CV review against this JD
  • Quest: a 5-day take-home (unpaid) where you build a reference, test a fast version against it, and lock the result in
  • HR alignment call (15–20 minutes)
  • 3-day paid test on a real task
  • Final evaluation by the Project Lead
  • Offer

  Requirements

  Required

  • Early to mid career: 1–4 years of hands‑on work in QA, test engineering, security testing or software engineering on real, live products (job, freelance or serious projects)
  • You believe that the better AI models get, the more quality is worth, and you see QA as your career leverage, not just a job
  • Security is your default: you check input handling, access, secrets and hostile input without being asked, and you can show one security problem you found
  • You have built an automated check that decides pass or fail with no human in the loop, and you can show it
  • You look for existing tools before building, and can say what you found and what gap was left
  • Daily, hands‑on use of AI tools in your work that you can demonstrate, including a place where you corrected what AI gave you
  • Clear written English: reports and handoffs that an engineer can act on without a call
  • You report numbers honestly and never claim a result you did not run

  Bonus

  • Experience with differential testing, property-based testing, fuzzing or a reference-implementation approach
  • Experience with image, file-format, graphics or media software
  • Experience testing mobile apps or Node.js backends
  • Experience on Web3 or crypto products

  Skills: AI-assisted QA, Security testing, Automated quality gates, Differential and property-based testing, Reference-implementation (oracle) testing, Root cause analysis, Python, TypeScript, Test automation, Written English reporting