AI HPC Storage Architect – System Design & Benchmarks

Socket.dev; Austin (TX); On-site; USD 150,000 - 190,000; Full time

Job summary

AMD is seeking an experienced HPC/AI storage architect to design reference architectures and deployment blueprints for AI infrastructure. You will work across compute, storage, networking, and orchestration layers to support internal teams, pre-sales engineers, and customers, with a focus on scalable storage solutions in AI/HPC clusters.

You will develop reproducible experiments, write design docs, and contribute to a growing library of technical artifacts.

Qualifications

• Experience with end-to-end HPC/AI storage architectures and reproducible benchmarks.
• Strong Linux fundamentals and performance tooling experience.
• Ability to translate complex systems into clear documentation and diagrams.

Responsibilities

• Shape AI infrastructure by creating reference architectures and deployment blueprints.
• Build a library of technical artifacts to support pre-sales engineers and customers.
• Evaluate HPC/AI stacks with focus on storage; document tradeoffs and fit.
• Design reproducible experiments and benchmarking harnesses; validate performance hypotheses.
• Present findings through demos, docs, and internal talks.

Skills

HPC systems thinking

Linux fundamentals

Performance profiling

Documentation and artifacts

Distributed training internals

Communication and writing

AMD ROCm ecosystem

Orchestration (Slurm, Kubernetes)

NCCL/RCCL

GPU/EPYC architecture

Education

Bachelor's or Master's in Electrical/Computer Engineering

Tools

Slurm

Kubernetes

Apptainer/Singularity

NDMA/Performance tools

Job description

AMD is seeking an experienced HPC/AI storage architect to design reference architectures and deployment blueprints for AI infrastructure. You will work across compute, storage, networking, and orchestration layers to support internal teams, pre-sales engineers, and customers, with a focus on scalable storage solutions in AI/HPC clusters.

You will develop reproducible experiments, write design docs, and contribute to a growing library of technical artifacts.