

Deepnote – A Better Data Science Notebook
ModelingDeepnoteNotebooksposted by ODSC Community November 5, 2020 ODSC Community

With the proliferation of data, notebooks gained popularity in both academia and industry as intuitive tools enabling code writing and execution, visualization, and insights sharing – all within one interface. Notebooks are now the go-to tool for data scientists for exploratory programming but come with their own set of challenges. Cumbersome versioning, reproducibility, lack of collaboration. Notebooks often encourage sloppy coding practices and they are notoriously hard to integrate with other systems.
As Jupyter users, we felt the pain of using data science notebooks every day. But we also knew how powerful they could be. So 2 years ago, we started to think about how to build a better data science notebook and created Deepnote. Deepnote is built on the top of Jupyter and extends the amazing work that has been done by the Jupyter community. We are using the same format as Jupyter and intend to remain fully compatible in both directions. However, in order to execute on our vision, we realized that with the Deepnote data science notebook, we must make some significant changes to how we think about notebooks today.
Embracing real-time collaboration in notebooks
Managing your data, models, and results as a data scientist is a tricky endeavor – even more so when working with others. Data science is a team discipline, but, data science notebooks generally don’t support collaboration.
With Deepnote, we decided to make collaboration a first-class citizen. To do this, we had to make some significant changes to the architecture. Deepnote runs in the cloud by default. We did this because we realized that the work of a data scientist is fundamentally different from the work of a software engineer. Data scientists don’t work alone — they share their work with others. Not just with other data scientists, but also with non-technical stakeholders. Every notebook is easily shareable, supports real-time editing and comments (just like Google Docs do).
Making data scientists more productive
Second, Deepnote makes you a better data scientist. We abstract away the complexity of setting up and managing hardware – Deepnote is built for the browser and platform-agnostic. Your projects in Deepnote are always available, with hardware up and running in a few seconds – only when you need it.
We’ve also redesigned the interface to encourage you to follow best practices, write clean code, define dependencies, and create reproducible notebooks. The autocomplete system and powerful command palette allow you to iterate faster, and the variable explorer enables you to identify insights in your datasets more easily.
Playing with the rest of your stack
Third, Deepnote integrates with other services. We don’t want to build just another data science platform where people work with an iframed notebook. We want to build an amazing notebook that plays well with other services, databases, ML platforms, and the Jupyter ecosystem.
Check out a 2-min demo here: https://www.loom.com/share/b7e05ecca78047c2a2f687d77be8ecea
Let’s re-invent notebooks together
Even though we started 2 years ago, building a new computational medium is a long journey. It’s both a hard technical problem and a user interface challenge. We’ve recently launched a public beta and have a lot in store for the future, including versioning, support for code reviews, full reproducibility, and visualizations.
If you’re a data scientist and like to try new things, we’d love for you to take Deepnote for a test drive.
If you’d like to chat with us about the challenges you face in using notebooks, we’d love to hear from you {email: team@deepnote.com}.