Databricks now lets developers create AI apps in 5 minutes: Here’s how
Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More
Databricks just made app development a piece of cake. The Ali Ghodsi-led company announced Databricks Apps, a capability that allows enterprise developers to quickly build production-ready data and AI applications in a matter of clicks.
Available in public preview today, the service provides users with a template-based experience, where they can connect relevant data and frameworks of choice into a fully functional app that could run within their respective Databricks environment.
According to the company, it can be used to create and deploy a secure app in as little as five minutes.
The announcement comes at a time when enterprises, despite being bullish on the potential of data-driven applications, continue to struggle with the operational hassle of the entire development cycle, right from provisioning the right infrastructure to ensuring security and access control of the developed app.
What to expect from Databricks Apps?
Much like Snowflake, Databricks has long provided its customers the ability to build apps powered by their data hosted on the company’s platform. Users can already build applications such as interactive dashboards to delve into specific insights or sophisticated AI-driven systems like chatbots or fraud detection programs.
However, no matter what one chooses to develop, the process of bringing a reliable app to production in a secure and governed manner is not an easy one.
The developers have to go beyond writing the app to handle several critical aspects of the development pipeline, right from provisioning and managing infrastructure and ensuring data governance and compliance to manually bolting integrations for access controls and defining who could use the app and who could not. This often makes the whole development process complex and time-consuming.
“App authors had to become familiar with container hosting technologies, implement single sign-on authentication, configure service principals and OAuth, and configure networking. The apps they created relied on integrations that were brittle and difficult to manage,” Shanku Niyogi, the VP of product management at Databricks, tells VentureBeat.
To change this, the company is now bringing everything to one place with the new Databricks Apps experience.
With this offering, all a user has to do is select a Python framework from a set of options (Streamlit/Dash/Gradio/Flask), a template of the type of app they want to develop (chatbot or data visualization app) and configure a few basic settings, including those for mapping resources (like data warehouses or LLMs) and defining permissions.
Once the basic setup is done, the app is deployed to the user’s Databricks environment, allowing them to use it themselves or share it with others in the team. When others log in, the app automatically prompts them with single sign-on authentication. Further, if needed, the developer will also get the option to customize the developed app and test their app code in their preferred IDE (integrated development environment).
On the backend, Niyogi explained, the service provisions serverless compute to run the app, ensuring not only faster deployment but also that the data does not leave the Databricks environment.
“Each app is fortified with robust security measures for seamless and secure user access. Plus, the integration with Unity Catalog provides comprehensive data governance and management capabilities, while the apps inherit the networking protections of your workspace, ensuring a multi-layered security approach for your sensitive data and applications,” he explained.
At this stage, Databricks Apps only supports Python frameworks. However, Niyogi noted that the company is working to expand to more tools, languages and frameworks, making secure app creation easier for everyone.
“We’ve started with Python, the #1 language for data. Anyone familiar with a Python framework can write their app in code, and anyone with an existing app can onboard it into Databricks Apps easily. We support any Python IDE. We are working with ISV partners to enable their tools to support Databricks Apps, and add support for other languages and frameworks,” he added.
Some 50 enterprises have already tested Databricks Apps in beta, including Addi, E.ON Digital Technology, SAE International, Plotly and Posit. With the public preview launching today, the number is expected to grow in the coming months.
Notably, Snowflake, Databricks’ biggest competitor, also has a low-code way to help enterprises develop and deploy data and AI apps.
However, Databricks claims to distinguish itself with a more flexible and interoperable approach.
“Databricks Apps supports Dash, Gradio, Flask, and Shiny as well as Streamlit, and supports more versions of Streamlit than Snowflake does. Developers can also use their choice of tools to build apps. We will continue to build on this flexible approach, adding support for more languages, frameworks and tools,” Niyogi pointed out.