GPU cluster marketplace for flexible AI compute access
San Francisco Compute operates a marketplace for buying and selling contracts on large-scale GPU clusters, enabling flexible short-term access without long-term commitments. It serves AI startups, labs, and enterprises needing on-demand high-performance compute for training models. This reduces financial risk in GPU infrastructure by creating liquidity in compute offtake contracts.
Build full-stack customer integrations and APIs to scale enterprise AI customers on a GPU offtake marketplace platform. Requires 3+ years of engineering experience, systems programming, SQL data modeling, and React/Next.js frontend work.
Owns hermetic, deterministic build system for TypeScript/Rust monorepo with protobuf integration. Migrates to Bazel/Buck2/Pants, implements remote execution/caching, ensures reproducibility across local/CI environments for scaling team.
Builds VM orchestration software for GPU neoclouds, provisions bare metal servers, creates Linux OS images, and designs APIs for marketplace users to configure compute resources. Requires strong systems programming skills in Rust or C, Linux familiarity, and knowledge of VMs/containers/syscalls.
Designs, architects, and scales production GPU/HPC clusters globally. Debugs hardware/software issues, automates operations, and mentors juniors. Requires 5+ years experience and hybrid SF presence.
Designs, deploys, and maintains high-performance networks for large-scale GPU clusters in HPC environments. Requires 10+ years experience with InfiniBand/RoCEv2 in CLOS topologies, automation, and hybrid work in San Francisco.
Builds compute delivery platform integrating GPU offtake market with orchestration software for HPC hardware. Requires systems programming (C/C++/Rust), SQL data modeling, gRPC/REST APIs, and control patterns expertise.