β€’

Drizz raises $2.7M in seed funding β€’

β€’

Featured on Forbes

β€’

Drizz raises $2.7M in seed funding β€’

β€’

Featured on Forbes

We find out before
your customers do.

We're a small team that tests AI for a living. We build simulated companies where an AI agent can fail safely, twenty times over, before it touches a real inbox, and we test mobile apps on real phones, which is where we started.

It started with a bug nobody caught.

In 2024, Asad, Yash and Partha kept meeting the same team: shipping a mobile app every week, with AI writing half the code and a QA process that hadn't changed since 2015. Bugs got to customers first. Everyone could see the problem; nobody wanted to be the one to fix it. So they did. Drizz became a Vision AI agent that reads a phone screen the way a person does, takes a test written in plain English, and runs it on real devices.
‍
Then the same teams started shipping AI agents into Gmail, Slack and Salesforce, and asked us to test those too. We tried, and learned something uncomfortable: an agent passes the demo and fails the twentieth run, and there is nowhere safe to find that out.

You can't point it at a real inbox. A mock doesn't push back. So we built the place that does: a simulated company, made of stateful twins of the tools, where an agent can break things that aren't real.

The result: one company that tests AI twice. The apps it ships, and the agents it runs.

Test suite dashboard showing 24 tests, 21 passed, 3 failed, with test names, status, device, and time.
Checkout flow test run details showing step trace with timings and a mobile order summary screenshot.

Fake companies. Real grades.

An agent that works in Gmail, Slack or Salesforce needs somewhere to practise. We give it a company that isn't real: twins of those tools that answer the same API calls, return the same errors, and reset in one call. The agent can't tell the difference. Your customers never see it.

You run a task twenty times from the same starting point. We grade what actually changed in each tool, not what the agent says it did, and hand you a pass rate, the ways it failed, and how two models compare on the same job.

Tata 1mg, Porter and NikahForever test their apps with us. A handful of AI-native teams are testing their agents with us too; we'll say who when they're ready.

Trusted by AI teams at

Built by people who give a damn.

Drizz started with three co-founders, Asad, Yash and Partha, who'd felt the testing pain firsthand and decided to do something about it. The team that's grown around them is small, young, and a little fearless: engineers who build twins of Gmail and Salesforce down to the error codes, and people who have watched more AI agents fail than almost anyone.

We hire people who are curious enough to break things and accountable enough to fix them. Mistakes aren't just tolerated here,Β  they're kind of the whole point. We fail fast, learn faster, and ship better for it every time.

Most of us are in our 20s. A few of us have scars from past workplaces. All of us are here because we'd rather build something that matters than something that's safe, and we'd rather do it together than anywhere else.

Group of people posing inside a restaurant with a red jeep-style table and colorful wall art.
Group of friends laughing and enjoying drinks at a lively outdoor restaurant at night.
Group of six young adults eating pizza in a modern office setting with open laptops in background.
Four men working on laptops at a shared office desk with monitors and backpacks.
Three men smiling indoors around a wooden table, one sitting on the table, two seated at it.

Ready to ship with
zero surprises?

Bring the agent you're building, or the app you're shipping. Thirty minutes with a founder; we run it and walk you through what we find.

Questions? We've got answers.

What is Drizz?

Drizz is an AI agent testing platform. Teams run their AI agents in a simulated company, built from stateful twins of Gmail, Slack, Google Calendar and other tools, twenty times from the same starting state, and get a graded report on every run before the agent reaches production. Drizz also tests mobile apps on real iOS and Android devices.

Who is behind Drizz?

Three co-founders, Asad Abrar, Yash Varyani and Partha Sarathi Mohanty, and a small team of engineers. The company was founded in 2024.

Do you still do mobile app testing?

Yes. It's where we started and it's still a product: describe a flow in plain English and Drizz runs it on real iPhones and Androids. Agent testing grew out of what we learned there.

Why "twins" and not mocks or sandboxes?

A mock returns whatever you told it to. A vendor sandbox is shared, slow to reset, and often gated behind a licence. A twin keeps state, enforces the real tool's rules, returns the real errors, and resets in one call, so the agent can't tell it apart from the real thing and nothing real gets touched.

Does any customer data enter a test?

No. You describe the company in plain language and we seed the twins from that. Real inboxes, calendars and CRMs are never connected.

Where does Drizz run?

In our cloud, or self-hosted in yours with Docker Compose or Kubernetes, including air-gapped.

Who are your investors?

Stellaris Venture Partners, Shastra VC and Better Capital.

Are you hiring?

Usually. Write to us at the address on the demo page and tell us what you've broken recently.