I first heard about DataPelago from Kushagra Vaid of Eclipse.vc in 2022. We were fellow panelists at a TSMC Symposium, where I mentioned I was searching for disruptive early-stage startups that could benefit from my experience.
Kushagra, whom I deeply respect, introduced me to Rajan.
Within an hour of hearing Rajan's vision, my entrepreneurial instincts were ignited: Rajan, a brain trust of Cavium, presented a vision to dramatically improve the cost-performance ration of data acceleration in the new cloud data era. This was at a time when AWS Nitro instances were revolutionizing the network and security offload in cloud services.
I have long believed that the scaling challenges faced by CPUs due to the limitations of Moore's Law will eventually impact data processing. The only way to breakthrough this barrier is to significantly enhance the efficiency of data processing through contextual data intelligence. While there have been many DPU accelerators available, they were limited to only handling network packet data and not raw enterprise data in the context of ETL. Moreover, existing accelerators certainly did not create software-based data execution environments at the query engine level. Given these constraints, I couldn't envision a way to achieve the orders-of-magnitude acceleration that is needed, regardless of the underlying hardware platform.
Rajan explained his plan to disrupt this by first building a software-based query plan that leverages inherent data knowledge to compute adjacency and storage to compute proximity. On top of this, he described his vision of creating networked query engines that could be used to achieve massively parallel query execution. Needless to say, I was convinced by Rajan’s vision and knew that it was time for me to get involved.
As I have observed throughout my career, the success of great ideas often depends on a team's ability to make subsequent adaptive executions of the original vision, while capitalizing on its core strengths at the time of inception.
DataPelago began with a vision to transform data acceleration with an intelligent layer driven by parallelized compute. From day one, DataPelago has cleverly architected their data execution abstraction to be processing unit agnostic, which now has expanded to include GPUs. This foresight now positions DataPelago as the undisputed leader in data acceleration for Data + AI. In this context, data acceleration is enabled by data intelligence on top of a parallel data execution infrastructure.
While an ambitious project like this carries significant risks, DataPelago has a visionary founder, a top-notch team, and a track record of proven results to support their claims at every stage. In DataPelago, I see a company founded by a new generation of entrepreneurs whose relentless pursuit of breakthroughs will guide us into the new Data + AI world. It has been a joyful journey with DataPelago and I am honored to be a part of it.