Data4Good Roundtable: TabPFN
Details
This time we’ll explore TabPFN, a foundation model for tabular data.
Our starting point is an episode of “Machine Learning Street Talk” podcast, in which co-founder of Prior Labs, discusses how TabPFN works, what makes it different, and what it could mean for the future of machine learning.
Machine Learning Street Talk - September 23, 2026 : How Deep Learning Finally Cracked Messy Tables – Frank Hutter
The episode is available on YouTube, Spotify, Apple Podcasts and other podcast providers.
This time, we’d also like to get a little hands-on! We’ll experiment with TabPFN together, try it on some tabular datasets, and discuss how it compares with more conventional approaches.
If you’d like to try it yourself, we recommend bringing a laptop and setting up a free Kaggle account beforehand. We’ll explore using Kaggle Notebooks so that no local installation or powerful hardware is needed.
As always, the focus is on exchanging ideas and exploring together, not on having all the answers. You’re very welcome to join even if you haven’t listened to the podcast or don’t plan to code along.
Who are we
We are CorrelAid, a network of more than 2500 data scientists who want to change the world with data science and machine learning. We are helping organizations that do good with data science projects. This is the Frankfurt Rhein-Main Local Chapter with about 30 people overall. Newcomers welcome!
This event is in English language and follows the CorrelAid code of conduct.
