Streamlit Inc., an artificial intelligence startup backed by Google LLC’s Gradient Ventures, today launched a service called Streamlit Sharing that’s designed to make it easier for developers to share their machine learning applications with users. 

Building a machine learning application requires more than just training a neural network. There also needs to be a user interface, to present processing results from the neural network, and a data ingestion pipeline that feeds the software the information it needs to process. 

San Francisco-based Streamlit offers a popular open-source tool of the same name that promises to reduce the task from weeks to hours. Despite having been released only a year ago, Streamlit already boasts more than a million downloads and several Fortune 500 users. The startup claims that developers have used its tool to build hundreds of thousands of machine learning applications to date.

Streamlit Sharing, the new cloud service launched today, is essentially a free hosting platform on which developers can run their Streamlit-built machine learning software. End-users can access applications hosted on the service via a browser. The catch is that Streamlit Sharing is only open to open-source applications whose code is publicly available on GitHub, but Streamlit is also building a paid offering for commercial projects that’s currently in private beta.

Streamlit Sharing aims to strengthen the open-source ecosystem around the startup’s tool. GitHub already provides a platform where developers can share machine learning projects, Streamlit Chief Executive Officer Adrien Treuille noted in a blog post, but there’s a technical barrier for end-users. Running a Streamlit application from GitHub requires downloading the raw code, running it and reading documentation. Applications hosted on Streamlit Sharing, in contrast, can be opened in a browser like any web service. 

“Github is overflowing with amazing ideas, models, algorithms, and datasets,” Treuille wrote. “But it’s just code, and code alone doesn’t let you play with models, see algorithms, or touch data. It’s too hard. You have to pip install something, import dependencies, and read through code examples all before you ever see it in action. What you want is a ‘play’ button.”

Streamlit Sharing lets developers upload their applications by pasting the link to the GitHub repository containing their code. By simplifying sharing of AI projects, the service lowers another adoption barrier to the startup’s tool, which should ultimately boost its user acquisition strategy. Having a large installed base of open-source users will provide useful market validation when Streamlit eventually launches its paid offering.

The launch of Streamlit Sharing comes about four months after the startup closed a $21 million funding round co-led by Google LLC’s Gradient Ventures fund and GGV Capital. 

Image: Streamlit