Rising to the challenge by leveraging collaboration and modularity were strong themes of Day 1 at JupyterCon 2025. Keynote speaker David Donoho spoke of data science as a continuous challenge with researchers both pushing the envelope and collaborating; papers are built on common datasets and ultimately share methods that anyone with the requisite expertise can use (Donoho, 2017). This reflected my formative experiences with data science; as an undergraduate completing an honors thesis, I modified a published computer vision workflow for identifying seabirds to instead identify seals by age and sex, leading to my own publication (Santini et al, 2025) which has in turn spurred other conversations about identifying marine mammals.
Now I’m the course software and infrastructure engineer for Data Science Undergraduate Studies here at UC Berkeley. My days are mostly improving the home grown software and technical workflows that enable the largest classes at Berkeley to run and helping our teaching assistants with the technical problems they encounter. I maintain things like course websites, the datascience package, and our publicly available textbooks and course notes. The undergraduate data science classes are built on the Jupyter ecosystem. Assignments are completed in Jupyter Notebooks on UC Berkeley’s JupyterHub (see Rebecca Dang’s presentation on the data engineering course). Our textbooks like Data 8 are deployed with Jupyter Book. And now you can work through the Build a Jupyter Book with The Turing Way tutorial yourself!
Collaboration is the culture of Jupyter. And this is amazing because collaboration aids learning at every level. Whether it’s high school students participating in Skew the Script, first year college students taking Data 8 at California community colleges or industry professionals, we all benefit from sharing knowledge and learning from each other. Talks highlight moves to open source previously proprietary software and a desire for partnership between open source and corporate.
Photo: UC Berkeley has a rich history of leadership in leveraging Jupyter tools for world leading data science education, and many educators and researchers - including our open source collaborators and Berkeley alumni - attended JupyterCon 2025
JupyterCon participants were asked to think about collaboration potentials with integrated AI agents. The Nomic and Deepnote keynotes both mentioned this, but perhaps it was best shown by Abigayle Mercer and Zach Sailer’s demo of real time collaboration with multiple AI agents. The agents stepped in without a chat interface, working directly in the notebook and even calling on each other when needed. Jupyter AI states their mission as “[creating] modular, extensible building blocks that enable humans and AI to collaborate effectively in Jupyter environments.” Plug and play with your desired models, personas, and create your own customized AI workflows.
I’m also excited about the collaboration potential from Jupyter Book 2, both to share resources directly and then to further leverage those resources in different contexts. Jupyter Book 2 is a modular and composable software that treats computation as a first class citizen. Jonathan Ferrari’s lightning talk brought Jupyter Books for open source education to the front (presentation slides). His live demo showed just how easy it is to contribute to a Jupyter Book that showcases higher education teaching modules: https://cal-icor.github.io/textbook/intro.html. Further, talks and tutorials demonstrated how artifacts created with Jupyter Book 2 can be utilized elsewhere. Obviously the notebooks and markdown files that make up the content of a Jupyter Book can easily be used in different contexts. But also you could take more complex features of a Jupyter Book available in the structured data, such as a glossaries, and use them in new ways.
Photo courtesy of Jenny Wong: Chris Holdgraf (2i2c) and Rowan Cockett (Curvenote) present Jupyter Book 2: Next-generation Tools for Creating Computational Narratives
I’m grateful for open source tooling that allows easy sharing of ideas and many different collaboration potentials. Jupyter empowers educators and students at UC Berkeley and elsewhere to communicate, learn, and collaborate. Every day, we use Jupyter tooling to deliver a scalable, reproducible, and more equitable computing environment and educational resources that meet the massive demand for data science skills.