A pre-trained relational foundation model can make accurate predictions on any database and any predictive task without any task-specific model training.

0:00The recent breakthrough uh that we had and we just released um in the second version uh is our what we call a relational foundation model. Um and that's a pre-trained foundation model uh that can reason over structured relational data. Um and it's crazy what this model can do. It can make accurate predictions on any database and any predictive task without any model training. All right, everyone. Welcome to another

0:48episode of the TwiML AI podcast. I am your host, Sam Cherington. Today I'm joined by Yuri Lecovitz. Yuri is co-founder and chief scientist at Kumo and a professor at Stanford University. Before we get going, be sure to hit that subscribe button wherever you're listening to today's show. Yuri, welcome to the podcast. It's great to finally connect with you. >> Yeah, great to be here. >> I'm looking forward to our chat. We're

1:14going to be digging into your work on relational learning um as well as some of the other interesting things you're up to at Stanford and and around AI for science and more. Uh but let's start there. Tell us a little bit about your research focus. >> Uh yeah, great. So uh I'm professor at Stanford here in the computer science department. Uh you know where the future happens I like to say. Um so there's

1:44always exciting research going on. Our u uh focus recently has been I would say on two areas. First is AI for science. Uh and in particular in we have a project that we call AI virtual cell where we are basically building next generation foundation models that allow us to represent human cells, patients as well as individual molecules in cells and allow us to re to reason um across