Python Jupyter notebooks are the lingua franca of modern data science. While Python has lots of great plotting and data visualization libraries, more can be done to integrate in-notebook visualizations by extending them to the Web, while retaining the ability to dynamically modify data renderings easily within Jupyter.

TileDB is motivated to improve this situation, both as the maintainer of the open-source TileDB Embedded library and its Python API, as well as a commercial service running hosted Python and R Jupyter notebooks for customers. In this talk, we present lessons learned working between Python, R, JavaScript, and even the TileDB data format itself — all in an effort to make Python Jupyter notebooks the centerpiece of data visualization workflows.

Supporting points:

• Automatically edit & publish interactive R Shiny dashboards straight from Jupyter notebooks.
• Adapting a 3D rendering engine for gaming (Babylon.js) for visualizing 3D point cloud data sets within Jupyter notebooks.
• Learnings from adapting Mapbox GL JS into hosted notebooks with TileDB arrays & vector tile server making TileDB Cloud arrays available to any web client that supports Mapbox Vector Tiles.
• Building a notebook versioning UI based on TileDB arrays.