The San Francisco Tensor Company focuses on reinventing the software and infrastructure stack for AI and high-performance computing. Their offerings include Emma Lang, a programming language for optimized computation, and Elastic Cloud, which manages infrastructure complexity across multiple hardware providers. The company aims to enhance performance and portability of code across platforms.
Automatically orchestrate training jobs across 1 to 10,000 GPUs; Transform kernels for optimal performance; Ensure code runs efficiently on GPUs, TPUs, and next-gen accelerators
Co-Founder
The San Francisco Tensor Company offers several key products and services focused on enhancing performance and portability in AI and high-performance computing (HPC). Their main offerings include:
Elastic Cloud: This service automatically identifies the most cost-effective hardware across various providers, orchestrating training jobs and managing infrastructure complexities, whether the user requires 1 or 10,000 GPUs. It simplifies the process of finding the cheapest hardware across providers.
Kernel Optimizer: This tool transforms computational kernels into their fastest mathematical forms by simulating memory, cache, and hardware topology, delivering performance that surpasses hand-tuned code. It effectively optimizes kernels for maximum efficiency.
Emma Lang: A new programming language designed for high-performance, hardware-aware computation, ensuring optimized code execution across diverse platforms, including GPUs and TPUs. Emma Lang guarantees that code runs efficiently across every platform.
These offerings are tailored for AI developers needing cross-cloud solutions and HPC teams managing large-scale compute resources.
Backed by Y Combinator; Elastic Cloud is now live; Automatic kernel optimization technology in development