Senior GPU Cluster Architect — Remote

Andromeda; United States; Remoto; USD 180.000 - 240.000; Tempo pieno

Vantaggi offerti da questo lavoro

Healthcare

Dental

Vision

401(k)

Unlimited PTO

Descrizione del lavoro

Andromeda is seeking a Cluster Architect to design and deploy large GPU clusters for AI labs and compute providers. You will convert customer requirements into complete cluster designs across compute, fabric, storage, management, and power/cooling, acting as the technical authority on builds.

You'll work with Head of Infrastructure across sites and vendors, delivering 1,000+ GPU deployments, optimizing for latency, bandwidth, and density while staying within site power constraints.

Competenze

• Experience designing 1,000+ GPU clusters.
• Deep knowledge of HGX/NVL72 platforms and reference architectures.
• Full stack familiarity: InfiniBand, Spectrum-X, Slurm or Kubernetes.
• Facility design: power, cooling, rack layouts.

Mansioni

• Create reference designs for current and next-gen NVIDIA platforms.
• Design customer configurations, BOMs, and rack layouts.
• Analyze workloads for latency, bandwidth, and GPU density.
• Design within fixed power budgets while maximizing density.
• Review vendor quotes and BOMs.
• Establish burn-in, performance and acceptance tests.
• Validate site power and cooling requirements.

Conoscenze

GPU clustering

Cluster design

Power & cooling design

Vendor coordination

Project leadership

Strumenti

InfiniBand

Spectrum-X fabrics

Slurm

Kubernetes

NVIDIA HGX

NVL72

WEKA

DDN

Descrizione del lavoro

Andromeda is seeking a Cluster Architect to design and deploy large GPU clusters for AI labs and compute providers. You will convert customer requirements into complete cluster designs across compute, fabric, storage, management, and power/cooling, acting as the technical authority on builds.

You'll work with Head of Infrastructure across sites and vendors, delivering 1,000+ GPU deployments, optimizing for latency, bandwidth, and density while staying within site power constraints.