Teams at Andromeda (andromeda.ai)

Owns reference and customer-specific designs for large-scale NVIDIA GPU clusters. Translates workloads and site constraints into compute, networking, storage, management, power, cooling, BOM, and rack-layout designs. Reviews vendor quotes, establishes acceptance criteria, validates data-center fit, and serves as technical authority across deployments. The role requires experience delivering 1,000-plus GPU clusters and deep knowledge of NVIDIA architectures, high-performance fabrics, parallel storage, and cluster management platforms.

Designs, deploys, validates, and troubleshoots large-scale GPU cluster networks, including InfiniBand, RoCE, Spectrum-X Ethernet, storage, front-end, and management fabrics. Responsibilities include topology and capacity planning, cabling and optics specifications, fabric bring-up, NCCL performance testing, deployment acceptance, incident escalation, and network automation using Python and Ansible.

Deploys large-scale GPU clusters from hardware delivery through production handover. Responsibilities include rack and stack, cabling, power-on, firmware management, burn-in, acceptance testing, vendor and contractor direction, quality inspections, punch-list closure, documentation, and deployment runbook creation. Requires production GPU deployment experience, Linux and hardware systems expertise, direct liquid-cooling commissioning experience, independent site ownership, and travel up to 50%.