Cedana

Cedana, based in New York and operating fully remotely, is a company in the B2B sector focused on infrastructure. With a team of four, Cedana participated in the Y-Combinator S23 batch and serves regions including the United States, America/Canada, and globally through remote operations. The company’s technology allows for real-time save, move, and restore of compute workloads, expanding GPU capacity and increasing CPU/GPU utilization by up to five times. This results in a reduction of latency by up to ten times and compute costs by 20%-70%, while also enhancing reliability and resource efficiency. Cedana’s solutions are applicable to AI training and inference, high-performance computing (HPC), data analysis, development tools, and infrastructure. The company aims to transform cloud compute into a real-time, arbitraged commodity and democratize access by removing barriers between builders and compute resources. Cedana’s system, which does not require code modification, automatically moves jobs across instances and vendors, maintaining progress. It also allows developers to set specific preferences for cost and performance, and assigns reliability and performance requirements across tasks. Cedana’s approach to vendor-agnostic compute arbitrage has been adopted by major entities like OpenAI, Microsoft Research, and Meta, addressing significant bottlenecks in access to compute resources and facilitating a global compute exchange.