Cluster Architect

Andromeda Cluster • San Francisco, CA

1d

About The Position

Andromeda’s infrastructure team designs, deploys, and operates large-scale GPU clusters for AI labs and compute providers. The Cluster Architect owns Andromeda’s reference designs, turns customer requirements into complete cluster designs spanning compute, fabric, storage, management, and power and cooling interfaces, and serves as the technical authority for each build. The position reports to the Head of Infrastructure.

The role covers reference designs for current and next-generation NVIDIA platforms, customer-specific configurations and rack layouts, workload analysis, power-constrained design, technical review of vendor quotes and bills of materials, cluster acceptance criteria, and validation of data center site requirements.

Requirements

• Designed and delivered GPU clusters of 1,000 or more GPUs.
• Detailed knowledge of NVIDIA HGX and NVL72 systems and SuperPOD reference architectures.
• Working knowledge of InfiniBand and Spectrum-X fabrics.
• Working knowledge of parallel storage systems such as VAST, WEKA, and DDN.
• Working knowledge of Slurm or Kubernetes cluster stacks.
• Understanding of rack power, liquid cooling, and how facility constraints shape cluster designs.

Nice To Haves

• Experience with NCCL, HPL, and MLPerf benchmarking and cluster acceptance.
• Experience delivering GB200 or GB300 NVL72 deployments.
• NVIDIA-certified architect or equivalent.

Responsibilities

• Create standard cluster reference designs for current and next-generation NVIDIA platforms.
• Develop customer-specific configurations, bills of materials, and rack layouts based on workload, scale, and site requirements.
• Analyze customer training and inference workloads to make design tradeoffs across latency, bandwidth, and GPU density.
• Design clusters within fixed site power budgets, maximizing GPU density while reserving headroom for storage, networking, and services.
• Technically review vendor quotes and bills of materials.
• Define burn-in, performance, and acceptance tests for determining when a cluster is ready for service.
• Validate power, cooling, and floor requirements against candidate data centers.

Benefits

• Comprehensive benefits for employees and their dependents, including healthcare, dental, and vision coverage.
• 401(k).
• Unlimited PTO.

What This Job Offers

• Job Type: Full-time
• Education Level: No Education Listed