Ankit Sharma

Qumin.ai - AI QA that writes and runs E2E tests in under a minute

by•
Qumin turns plain-English requirements into self-healing end-to-end tests for web and Android ( soon ). No brittle selectors, no test maintenance, just describe what to test and Qumin generates, runs, and repairs it automatically. QA in under a minute.

Add a comment

Replies

Best
Ankit Sharma
Hunter
šŸ”Œ Plugged in
šŸ“Œ
Happy to introduce Qumin today. QA has been the same story for a decade — write tests, tests break when UI changes, engineers waste hours fixing selectors instead of shipping. Khushal and team are attacking this with vision-language models that understand what the app looks like, not just what the DOM says. Self-healing E2E tests from plain English for Web and Android ( soon ). Have been tracking what @khushal_bapna is building for a while — sharp team, real problem. Excited to see what the PH community thinks. Makers are here to answer questions — over to you @khushal_bapna @srihari_maruthachalam šŸ‘‹
khushal bapna

Hey Product Hunt šŸ‘‹ I'm Khushal, Founder of Qumin.

We built this because every QA team we talked to was drowning in the same problem: tests break every time the UI shifts. You spend more time fixing tests than writing them.

Qumin uses vision-language models to understand what your app looks like and does — not just CSS selectors. You describe a test in plain English ("log in, add item to cart, verify total"), and Qumin writes it, runs it, and self-heals when the UI changes. Works across web and Android ( soon ).

Under the hood: multi-modal LLMs for plan generation and executions.

Would love your feedback — especially teardowns.

If you're running E2E tests today: what's the last test that broke on you, and what was the actual root cause? Curious to see if Qumin would've caught it.


— Khushal

Srihari Maruthachalam

Hi Product Hunters,

Qumin does one thing exceptionally well: it tests your app the way a human would, not the way a DOM parser does.

Our agent looks at the actual rendered UI — buttons, layouts, state — using vision-language models, figures out the steps, executes them, and heals itself when things shift.

No selectors. No XPath. No 3 am "element not found" pages.

That's the core bet: if the agent sees what the user sees, most flakiness disappears by design.

I'm Srihari, co-founder at Qumin. Hunt Qumin.ai to know more.