How Knowunity Ships 3x Faster with Flutter, Gemini and BigQuery

Google recently visited Knowunity for a developer story about how a small team builds the app that 8 to 10 million students across 26 countries use every month. The goal we set ourselves is a big one: the number one learning companion for one billion students globally.

How a small team builds all this

Knowunity was built on Flutter from the very first line of code in 2020, back when we were a single mobile developer. Instead of writing the app twice, once for iOS and once for Android, Flutter lets us write it once and ship it everywhere. We crossed 50,000 users after two weeks and one million after a year, and every new feature reached every student on the same day. Dart is quick to pick up, too, so web, Android and iOS engineers all become productive in the codebase within weeks.

You may have noticed that Knowunity moves. Screens slide into each other, a correct answer lands with a little celebration, and when you earn enough experience points to level up, Knowie flies across the screen to cheer with you. That is deliberate: we spend a lot of effort on engagement, and animation is a big part of it. We combine Flutter's built-in transitions with custom animation files built in Rive, so those moments look like nothing else out there.

Your questions do not all go to the same model either. We use a range of Gemini models and pick the right one for the job: a light, fast model like Gemini 3.1 Flash-Lite for simple tasks where you just want a quick answer, and stronger reasoning models when a problem is genuinely complex. That is how we keep answers both fast and high quality.

AI has changed how we work as much as what we ship. We now run A/B tests and product improvements about three times faster than before, and it takes us half the time to fix a bug. An engineer hands a design file or a specification to an AI agent, the agent implements it and tests it end to end inside the Flutter app, and the engineer reviews the result on their own device before approving it.

Built for students everywhere

Knowunity started in Berlin, but most of the students using it today are somewhere else. Once we moved into new countries, translating the app properly became essential. Flutter's localization tooling and ARB files let us localize every piece of content in the app, which is how the same learning companion feels native whether you open it in Berlin, Bogotá or Manila.

Underneath, Google BigQuery is our data warehouse. Frontend analytics, backend data and every AI request from all 26 countries come together in one place, so we can see what actually helps students learn, and so our own AI agents can spot patterns and suggest what to build next.

Where this is going

It would have been a lot harder to grow to that size and that scale without Flutter.

Knowunity has always been meant to be a daily learning companion: personal to how you study, and honestly a bit more fun than the alternatives. Flutter, Gemini and BigQuery are the tools; a billion students learning smarter is the point.

Thanks to the Google Flutter team for telling the story with us. Watch the full video on YouTube.