Kumo AI+Relational Workshop at AWS

On Wednesday I’ve attended a workshop at AWS hosted by Kumo. The founder and CEO, Jure Lescovec, a legendary graph guy, lead it himself. Jure gave a talk on Graph Transformers. They are solving the positional encoding for graphs, bringing all the machinery we have to bear on graphs.

The graphs are composed from customer tables. We had a hands-on experience running Kumo on some generic data with customers, stocks, and trades. It was really easy to identify primary keys linking tables and formatting the fields for graphs. A graph is then composed by the system and stays separate from the original tables until you ask for an update, if you update the tables.

The prediction works with a SQL-like language, with syntax and field completion. Kumo team was present tin force. Hema Raghavan, Kumo cofounder, had a fireside chat with Ben Shahshahani, the Chief AI Officer at Cleveland Clinic. The prospects of graph-based predictions in healthcare are enormous.

Many Kumo engineers helped the participants. Manan Shah, a founding engineer, was especially helpful to me.

Here’s the full f.photo gallery from the workshop.

Overall, what a great team, product, and potential!