A former Google X employee just raised $21 million for his AI startup Streamlit, which is already being used by companies like Uber and Stitch Fix

• On Tuesday, the artificial intelligence startup Streamlit announced it raised $21 million in Series A funding.
• Streamlit CEO and co-founder Adrien Treuille first saw snags in building AI apps when he worked on projects like self-driving cars and Google Glass at Alphabet research subsidiary Google X, now known as X.
• With the funding, Streamlit plans to invest in its open source community and build its paid enterprise product.

While still at Alphabet's research subsidiary Google X (now known as X), Streamlit CEO and co-founder Adrien Treuille worked on all sorts of projects, from self-driving cars to Google Glass.

But during this time, he first started to see a problem: Engineers building artificial intelligence products like a self-driving car constantly faced ever-growing mountains of data, and they often had difficulty working with all that.

He started building software to address the bottleneck, making it easier for data scientists and developers to build AI apps and experiment with machine learning, a field of AI that enables computer programs to learn from data, identify patterns, and make decisions on their own, without being told what to do. Machine learning is used in self-driving cars, email spam filters, mapping, and more.

What started as a personal project ended up getting used by Uber and Stitch Fix. The next thing Treuille knew, investors wanted to put money into it. And on Tuesday, Streamlit announced it raised $21 million in Series A funding led by Gradient Ventures and GGV Capital.

With the funding, Streamlit plans to invest in its developer community and build out its enterprise product, which would include features for security, logging, and scalability.

Right now, its software is available as open source, meaning it's free for anyone to use, download, or modify. The startup's paid enterprise product hasn't launched yet, but already, there's a list of more than 1,500 companies waiting to use it, Treuille says.

"We're trying to build something lasting and generational," Treuille told Business Insider, pointing to the open source databases MongoDB and Redis as examples.

Streamlit's growth skyrocketed since it first launched seven months ago

Treuille was previously a project lead at Google's research subsidiary Google X and a computer science professor at Carnegie Mellon University who built science video games. After more than three years at Google X and a stint at the self-driving car company Zoox, Treuille decided to take some time off.

During this time, he traveled around the world with his fiance and his child. He also started working on machine learning projects.

"I wanted to program more actually," Treuille said. "The problem is you stop doing the fun things. That was the core of it all."

Just seven months ago, Streamlit made its software available as open source. Now, it has been used to build more than 200,000 applications. It started growing quickly and globally as well, because people would volunteer to translate documentation to other languages. It's also compatible with popular open source projects like TensorFlow and PyTorch.

"Fast forward to the fall when we release it, it starts growing way beyond our expectations, at least in terms of the speed and of our growth," Treuille said. "People just totally got it and were enthusiastic about it. At this point, it looks like it's a legitimate, early stage, open source movement."

An 'almost ridiculous David and Goliath move'

Treuille says that one mistake Streamlit made early on was marketing itself as a competitor to the Jupyter Notebook, a popular open source app that allows people to create and share documents with live code, visualizations, and more. This was a mistake, Treuille says.

"We realized that Jupyter Notebooks was a massively well-entrenched technology around the world," Treuille said. "Going up against it was an almost ridiculous David and Goliath move. That was a moment when we knew we were onto something but we also felt we were heading 60 miles per hour into a brick wall if we tried to execute on this strategy."

The team realized that Jupyter Notebook wasn't a competitor at all. While it helps people analyze data, Streamlit was for building apps.

"That's when we started to feel the rush of wind against our faces," Treuille said. "Since then, in many respects, that was a fateful moment of course crashing."

Got a tip? Contact this reporter via email at rmchan@businessinsider.com, Signal at 646.376.6106, Telegram at @rosaliechan, or Twitter DM at @rosaliechan17. (PR pitches by email only, please.) Other types of secure messaging available upon request.