Kumo, a startup offering an AI-powered platform to solve predictive problems in enterprises, today announced that it has raised $18 million in a Series B funding round led by Sequoia, with the participation from A Capital, SV Angel and several angel investors. Co-founder and CEO Vanja Josifovski said the new funding will go towards Kumo’s hiring and R&D efforts on the startup’s platform and services, which include data preparation, analytics data and model management.
Kumo’s platform works specifically with graphical neural networks, a class of AI system for processing data that can be represented as a series of graphs. Graphs in this context refer to mathematical constructs consisting of peaks (also called nodes) which are connected by edges (or lines). Graphs can be used to model relationships and processes in social, computational, and even biological systems. For example, the link structure of a website can be represented by a graph where the vertices represent the web pages and the edges represent the links from one page to another.
Graph neural networks have powerful predictive capabilities. At Pinterest and LinkedIn, they’re used to recommend posts, people, and more to hundreds of millions of active users. But as Josifovski notes, they are computationally expensive to operate, making them prohibitively expensive for most businesses.
“Many companies today trying to experiment with graphical neural networks have not been able to scale beyond training datasets that fit in a single accelerator (memory in a single GPU), significantly limiting their ability to take advantage of these emerging algorithmic approaches,” he said. TechCrunch in an email interview. “Through fundamental infrastructural and algorithmic advancements, we have been able to scale to multi-terabyte datasets, allowing graph neural networks to be applied to customers with larger and more complex enterprise graphs, such as social networks and multi-sided markets.
Using Kumo, customers can connect data sources to create a graphical neural network that can then be queried in a structured query language (SQL). Under the hood, the platform automatically trains the neural network system, assessing its accuracy and preparing it for production deployment.
Josifovski says Kumo can be used for applications such as new customer acquisition, customer retention and retention, personalization and next best action, abuse detection, and financial crime detection. Previously CTO of Pinterest and Airbnb Homes, Josifovski worked with fellow Kumo co-founders, former Pinterest chief scientist Jure Leskovec and Hema Raghavan, to develop graph technology through Stanford University research labs. and Dortmund.
“Companies spend millions of dollars storing terabytes of data, but can only effectively leverage a fraction of it to generate the predictions they need to make forward-looking business decisions. The reason for this is the major gaps in data science capability, as well as the significant time and effort required to successfully bring forecasts into production,” Josifovski said. “We’re enabling businesses to move to a paradigm where predictive analytics goes from a scarce resource used sparingly to one where it’s as simple as writing an SQL query, enabling predictions to become fundamentally ubiquitous. – much more broadly scaled to use cases across the enterprise in a much shorter time frame.
Kumo remains in the pilot stage, but Josifovski says he has “over a dozen” early adopters in the company. To date, the startup has raised $37 million in capital.