HPC/GPU Cluster Architect

Founding HPC / GPU Cluster Architect

Remote, US

I'm working with a GPU infrastructure company building and operating supercomputers, with a market that lets customers resell reserved compute they don't use.

They build the data centers, GPU clusters and cloud infrastructure behind the service. Customers can access VMs, bare metal or managed Slurm, with unused reserved capacity sold back through the market to reduce idle costs.

The company has raised approximately $52M, with its latest Series A announced at a $300M valuation. It has also publicly announced $245M in compute contracts over 36 months across two B300 clusters, creating substantial demand for new infrastructure. The team is around 60 people, with plans to double.

This is a founding, high-ownership hardware architecture role for someone who wants to shape how the next generation of GPU clusters is designed.

You’ll work on:

• Turning loosely scoped customer compute requirements into cluster architectures and bills of materials
• Sizing GPUs, servers and networking around customer workloads
• Making hardware decisions across NVIDIA platforms, fabrics, power and cooling
• Designing and validating network topology alongside a dedicated network architect
• Evaluating current and next-generation NVIDIA systems and AI infrastructure
• Working directly with the CTO and founding architecture team to define technical direction
• Mentoring junior engineers and supporting exceptional hardware escalations, with dedicated teams handling deployment and routine operations

Requirements:

Personal ownership of architecture and delivery for a large, interconnected production GPU cluster

Clear understanding of how the GPUs were networked and operated together; experience with 1,000+ GPUs is preferred

Experience translating customer workload requirements into hardware sizing, component choices and bills of materials

Strong knowledge of current and next-generation NVIDIA GPUs and the AI stack

Ability to understand and independently validate cluster network topology

Practical depth across GPU servers, fabrics, power and thermal constraints

Independent senior-level delivery, strong technical judgment and experience mentoring engineers

Willingness to travel to San Francisco quarterly, attend additional architecture sessions as needed and support rare major-incident escalations

Experience with Slurm, Kubernetes, Linux, drivers, RDMA, virtualization or data-center telemetry is highly valuable

Package: $220K-$400K+ base + equity.