Partnering with Kumo: Predictive AI for All

We are thrilled to support Hema, Jure and Vanja on their mission to make graph learning easy to use.

By Konstantine Buhler

Published April 7, 2022

Kumo founders Hema Raghavan, Jure Leskovec and Vanja Josifovski.

When I met Jure Leskovec eight years ago in a Stanford classroom, it was immediately clear that he was special. I was an AI grad student taking his CS 246: Mining Massive Datasets course, and he was the charismatic star professor. As brilliant a communicator as he was a researcher, he had the kind of energy and curiosity that made students actually want to ask questions.

I’d also heard wonderful things about Vanja Josifovski, Jure’s co-founder at Kosei, and I watched with great interest when Pinterest acquired their company the following year and they stepped into roles as the team’s new Chief Scientist (Jure) and CTO (Vanja). Together, they went on to develop cutting-edge AI, help Pinterest grow into a massive success, and cement themselves among the best graph neural networks scientists in the world.

When the two of them reached out to Sequoia in late 2020 to talk about some new ideas, we were eager to do so.

None of us wanted a solution in search of a problem—so the question was, could we identify the right opportunity? Jure and Vanja’s goal was to build a company that would endure, and they were willing to be patient. Over the next six months, we talked through a dozen possibilities.

At the same time, we were looking for the right co-founder to add to this already all-star team—someone with the rare combination of deep technical knowledge in AI and experience building large engineering organizations. We kept hearing rave reviews about Hema Raghavan, who was leading Growth AI at LinkedIn, and once we met her, it was a no-brainer. She is nothing short of amazing both technically and culturally, and we are fortunate that she agreed to join us on this journey.

As our conversations continued, we kept coming back to the fact that companies spend untold billions storing their data in warehouses and lakes, all with the hope of making better business decisions. Yet today, those decisions are possible only for a select few enterprises who can afford to recruit and retain AI teams—and even those teams are constantly overwhelmed with work and tool sprawl. As experienced AI leaders themselves, all three of Kumo’s co-founders understand this firsthand.

Of course, data warehouses themselves once faced similar issues—until Snowflake came along. What if there were a platform that was just as accessible and easy to use, but for graph-based AI, so a company could load it up and immediately start making faster, smarter predictions? It was an exciting idea, and we knew it would be amazing—if it were technically possible. What the co-founders were proposing was an extremely difficult feat of engineering. 

But another year and a team of 20 later, Jure, Vanja and Hema are making that vision reality. Some of the best graph neural networks engineers out there have come together to bring AI not just to one company but to everyone. As Kumo continues to grow the team, scale its tech and invest in R&D, this Series A round is just the latest milestone in what we at Sequoia believe will be a decades-long partnership.

We’re at the precipice of a new wave of innovation in AI. CDOs, Data Engineers, Data Scientists, and Data Analysts alike are already recognizing what this “20-person AI team in a box” can do for their organizations, and—just as we’d dreamed way back when Kumo began—one Fortune 500 data leader recently told me, “This reminds me of when we first spoke with Snowflake.”