Feldera: Data processing built for speed and scale of AI

When Fortune 100 and tech native public companies with some of the largest and most complex data pipelines in the world trust a tiny startup with their critical production workloads, it must be doing something special. That is Feldera in a nutshell. Its customers are using Feldera to power agent identities, fraud models, supply chain, customer analytics, and more.  

With Feldera, enterprises get extraordinary improvements in latency and cost. We’re not talking about a 30% cost reduction or a 25% speed increase. We’re talking about orders of magnitude gains along both dimensions for large-scale data pipelines. Even though customers understand the theory, they are still stunned when they see the actual results on their own data. Those are the types of improvements that not only optimize existing workloads but allow companies to do what previously was not possible. Models can be served with up-to-date data. Agents can make decisions on data that can actually keep up with them.

When we invested in Feldera a couple of years ago, it was a research paper that had been proven out in a limited way at VMware. But we were captivated by the founders’ vision to flip the data processing model on its head. We believed they could do this because they were not industry insiders. They were not tethered to existing technology. Their backgrounds were in networking, operating systems, distributed systems, and compilers. But their outsider perspective gave them a very different framework for solving the big, age-old problem of processing data at scale.

The insight was simple - rather than recompute the entire data set from scratch each time they had to do a computation, they just computed the changes. Incremental computation was not new but no one had built a product that was usable, maintainable, and extremely flexible. To solve for this, they devised a framework for incremental computation that enabled the use of fully expressive SQL. This was the research that won best paper at VLDB.

Lalith, who is founder and CEO, has led the team from a lab to a company that now runs data pipelines for the most demanding enterprise customers. And they are just getting started. 

Congratulations to Lalith, Leonid, Mihai, Ben, Gerd, and the entire Feldera team on their Series A. We’re thrilled to have backed them from inception.