Recently, ML startup Streamlit raised $35 million in a Series B round led by Sequoia. Streamlit co-founders Amanda Kelly, Thiago Teixeira and Adrien Treuille launched the company in 2019 to make app creation as easy as Python scripting. “We have obsessed over making this alchemy — from scripts to apps — as fast and joyful as possible,” wrote Treuille in a blog post.
Streamlit: The Standard-Bearer
Streamlit was started to take data to another level by leveraging AI and machine learning. Streamlit wants data scientists to build predictive models in a novel way. Streamlit offers workflows that help build apps ranging from advanced analytics dashboards to sales and marketing tools based on the latest predictive algorithms. According to Streamlit, the platform’s unique workflow is many times faster than other Ml app building services. Data scientists can go from ideation to deployment in a matter of a few hours. Today, Streamlit’s users include Apple, Ford and Uber. It has more than 14,000 GitHub stars, and has been downloaded nearly two million times already.
“Streamlit apps are simple interactive script visualisations – a deceptively powerful idiom that strikes just the right balance between low code, power and customisability. This unique approach enables such fast creation of powerful, useful apps, that Streamlit apps have become an entirely new workflow within companies — similar to Google Docs and Notion. Streamlit for Teams lets companies instantly bring these apps into the entire company, allowing everyone to make faster, data-informed decisions,” said Adrien Treuille, co-founder and CEO of Streamlit.
Streamlit’s open-source app framework empowers data scientists. With the introduction of Streamlit for Teams, Streamlit has unlocked data scientists’ ability to instantly deploy and share Streamlit apps with teammates, clients, and other stakeholders. For instance, Streamlit for Teams is a zero-effort cloud platform to deploy, manage, debug and collaborate on apps securely. The product deploys apps directly from private Git repos and runs continuous integration to update apps on commits instantly. It layers on enterprise-grade data security and OAuth2-based authentication as well as advanced collaboration features for both data scientists and their customers.
According to Treuille, machine learning engineers might get stuck with time-consuming tasks such as building better tools to understand data. Since ML personnel are trained in niche domains, it is quite possible that building new tools will eat into their workloads. This is where Streamlit comes in handy. It enables data science teams to build these tools in a faster and in more non-technical way.
Growing interests of investors in companies like Streamlit, indicates the rising demand for deployment-as-a-service. The algorithms are ready, and so is the hardware. So, why take the headache of building tools from scratch? Wouldn’t it be more productive to concentrate on data preparation than building tools? For instance, Streamlit enables machine learning engineers to build tools with only a few lines of code. These tools can further be used to understand the data and visualise it better; for example, building a set of sliders with different variables to interact with the data or simply creating tables with subsets of data that make sense to the engineer.
“Streamlit’s toolset has the potential to dramatically transform the way machine learning engineers work with the data in their models.”
Streamlit is compatible with atex, OpenCV, Vega-Lite, seaborn, PyTorch, NumPy, Altair, and other such libraries.
Here are a few additional advantages (source: Terence Shin):
- Python friendly.
- No need to know HTML and CSS.
- Covers UI components such as checkbox, slider, a collapsible sidebar, radio buttons, file upload, progress bar, etc.
- It supports multiple interactive visualisation libraries such Latex, OpenCV, Vega-Lite, etc.
Startups can take a lesson or two from Streamlit’s success and offer AI/ML teams tools to build tools, and seamlessly integrate them to their projects. Write code, save it, review the output, write some more, and so on, until you’re happy with the results. Streamlit was founded to create an interactive app for reviewing, debugging and sharing the code.
Data scientists should be able to visualise, mutate, and share data with custom made APIs — be it for displaying data or optimising performance. For example, in the case of Streamlit–as long as an app is running– every time a new element is added to the script and saved, Streamlit’s UI will ask if one wants to rerun the app and view the changes. Customisations like these are at the heart of many successful ML startups today.