Engineering Jobs at SF Tensor

SF Tensor provides an optimized, cross-vendor training stack for enterprise post-training and frontier pre-training. The company helps model builders train models by compiling kernels, finding capable hardware across providers, and operating training runs from 1 GPU to 10,000 GPUs. For enterprises, it enables continuous post-training on proprietary domain data without handing over the data or assembling an ML infrastructure team, with deployment options in the enterprise's cloud or SF Tensor's cloud. For AI labs, it supports frontier-scale pre-training of novel architectures with cross-vendor portability.

Its roles are in San Francisco, CA, US (6).

You own whole parts of the product rather than one service inside a large system. Expect to write the code, deploy it, and get paged when it breaks. Reviews are quick and the codebase is small, so a decision made on Tuesday can ship on Thursday.