AI Product QA Engineer
This is not the QA you know. We're not hiring someone to find bugs. We're hiring someone to guard the user's experience — with AI as the foundation.
Read this before you read anything else
Forget everything traditional QA taught you.
I don't want "I tested it, here are the 14 defects." I don't want a gatekeeper at the end of the line ticking checkboxes against a technical feature list. That QA is dead here, and honestly, AI already does it better.
I want business-product QA. Product taste as a discipline. User empathy as a test case. Your foundation is AI — everything runs through Claude Code / Codex — and on top of that foundation you apply QA from a business, user-empathy, onboarding, and go-to-market lens. You're not asking "does the button work?" You're asking "is this the sharpest user journey this product could possibly have?"
We're extreme about four things and nothing else: extreme hard work, extreme respect, extreme shipping, extreme innovation. Not hours . There's no boss — there's a product, a user, and you. You're rewarded on one thing: your hunger.
The 90 / 10
Same philosophy that runs this whole company. 90% of the work, AI does. Claude Code writes the test cases, automates the suites, drives the API calls, spins up the harness — faster than any QA team I've ever run . That 90% is the floor now.
The 10% is you — the human who decides what's actually worth testing and why:
- the user empathy — feeling the friction a real brand manager feels at 9pm
- the judgment — is this the sharpest journey, or just a working one?
- the taste — the design standard, the elegance, the delight
- the business sense — does this survive a real client's Tuesday, their onboarding, their objection?
- the critical eye — reading an AI trace and knowing instantly when the product is confidently wrong
AI didn't shrink QA. It concentrated it. The 10% is smaller and heavier than the 100% ever was. That's the 10% I'm hiring.
How you actually work
Everything you do runs through Claude Code. You write the cases in it. You automate the cases in it. You do business testing in it. You interrogate user empathy in it — "how will the user react to this, and is this the sharpest journey of the product?"
But make no mistake: this is technical QA. You write scripts. You hit and test APIs directly. You read the traces, the payloads, the failure modes. You don't wait for a UI to click — you go straight at the system. The difference from old-school QA isn't less technical; it's technical in service of the user and the business, not the feature checklist.
The stack & the surface
We're a web application in Python, Rust, and everything in between — and you'll be QA-ing across four products: ARIA, the Persona product, the video-decoding product, and the rest of the family. Real APIs, real pipelines, real agents watching millions of social videos in Hindi, Tamil, Bahasa, Thai for the world's biggest consumer brands . You need to be comfortable writing a script to probe any of them and reading exactly what came back.
Your first 48 hours
Day one, morning: you get access to a live product and its real client questions in the queue. Not a sandbox. Not a tutorial.
Day one, afternoon: you pick one user journey and, in Claude Code, you build the case, automate it, and pressure-test it against the actual experience a client would have.
Day two: you tell me not "here are the defects" but "here's where the journey breaks the user, and here's the sharper one." And you've already got the automated eval that guards it going forward.
If that scares you, this isn't your room. If it makes you grin — keep reading.
The loop
We are always on. A technique drops on Twitter Tuesday morning; by Thursday it's in a product; by Friday a brand team on the other side of the world is using it — and someone has to make sure that journey is delightful, not just functional. That someone is you. See it. Test it against the user. Sharpen it. Ship it. The loop never stops. If your learning lives in "read later" folders , this will break you.
What you own
The user's experience across every product. Not the defect list — the journey. Find a problem and you don't just log the instance; you build the automated eval in Claude Code that kills the whole class of bad experience forever. Your evals become the product's conscience. Your standard becomes the bar the whole team ships to.
The bar is craft, not years. 23 or 43 — I don't care. I care whether you can feel what a user feels and write the script that proves it.
Who this is actually for
This is a close-to-100% self-starter environment.
You do not want to be here if you want to be handed a test plan. You want to be here if you believe you're the only one who can guard this product's soul — the king or queen of the user journey — with absolute hunger burning in you to do it. Here, will is bigger than skill. And AI isn't a tool you use; it's the foundation you think on.
Fair warning ⚠
The pace is relentless and there's no playbook for AI-product QA — you're partly inventing the discipline as you go. If you want a stable checklist and a comfortable gate at the end of the pipeline, you'll be miserable here, and I'd rather you know now. But if you want to define what QA even means in the age of agents — what you learn here in one year, you won't learn anywhere else in five.
How this goes
Don't send a resume. Send the 10%.
An eval harness you're proud of. A test suite you built in Claude Code that caught something a checklist never would. A teardown of a product's user journey where your taste and your technical chops were unmistakably yours.
Pay: ₹1,513,614.86 - ₹2,382,125.68 per year
Work Location: Remote
