Best Practice for Collaborating with Jupyter Notebooks

When we started an new NLP project, using python and Jupyter notebooks seemed like a no-brainer, but soon started running into challenges when we tried to share these notebooks using Github. After several disgusting merge conflicts, broken notebooks, and "it runs on my machine", we realised that we need to come up with some better ways of working to make our notebooks more practical.

This talk breaks down the simple and practical steps that we took to overcome the challenges of using Jupyter notebooks in a collaborative team. We outline the steps that we took to turn our notebooks from scrappy, conflicted and broken fragments of code to a high-powered collaborative tool that was easy to transform into a deployable python program down the line.

The result was a streamlined git process, practical and usable library to building out our ideas and the same high-speed, intuitive exploration of data that Jupyter notebooks are well known for.