Lightning Talk: GitHub-First Jupyter: How Pinterest Solved Notebook Collabo... N. Kagita & C.W. Chen
Lightning Talk: GitHub-First Jupyter: How Pinterest Solved Notebook Collaboration at Scale - Nikhil Kagita & Chia-Wei Chen, Pinterest
Learn how Pinterest built a comprehensive GitHub integration for Jupyter, transforming notebook development from isolated experiments into collaborative, version-controlled workflows.
This presentation covers our production system serving hundreds of data scientists, engineers, and analysts. We'll dive deep into a key challenge: balancing the power of Git with the need for a user-friendly experience for a diverse user base.
Discover how we created a system that provides the benefits of version control through both a GUI and simplified commands (like `git-safe-push`), while abstracting away complexities. We will also cover practical hurdles in enabling a GitHub-based workflow, including our solutions for blocking code reviews for PII, managing service account permissions, and streamlining pull requests.
Key topics include architecture for GitHub-backed Jupyter deployments at scale, tailoring the Git experience for users, navigating automated PII scanning and service account permissions, implementing custom pull request workflows, creating executable/version-controlled reports, and migration strategies from legacy Jupyter environments.