Overview

SF Tensor builds an AI/HPC software and infrastructure stack to reduce the infrastructure burden for AI teams. It offers Emma Lang, a hardware-agnostic programming language that runs across GPUs and TPUs without rewriting code, and the SF Tensor Stack, including Kernel Optimizer and Elastic Cloud. The Kernel Optimizer turns models into efficient mathematical forms by simulating hardware topology, often outperforming hand-tuned code, while Elastic Cloud finds cost-effective hardware across clouds and coordinates large-scale training. The goal is to remove vendor lock-in, enable cross-cloud, hardware-agnostic compute, and cut compute costs so AI researchers can focus on innovation.

YC Company

Launched Recently