Member of Technical Staff - GPU Infrastructure

Prime Intellect; USA; Hybrid; USD 150.000 - 300.000; Vollzeit

Benefits dieser Stelle

Remote work option

Equity incentives

Professional development budget

Team off-sites & conferences

Visa sponsorship & relocation support

Zusammenfassung

Prime Intellect is looking for a Role overview expert to design and deploy GPU infrastructure for large-scale AI training, inference, and HPC workloads. This customer-facing role blends cluster architecture with hands-on implementation, performance tuning, and on-call support across hardware and software.

You will translate workload needs into GPU cluster architectures, optimize production clusters with SLURM and Kubernetes, and contribute to capacity planning for deployments spanning hundreds

Qualifikationen

• Hands-on experience with GPU clusters and HPC environments (3+ years).
• Production experience with SLURM and Kubernetes; InfiniBand configuration/troubleshooting.
• Strong knowledge of NVIDIA GPUs, CUDA, drivers, and GPU container runtimes.
• Experience with infrastructure automation (Ansible, Terraform) and scripting (Python, Bash).
• Knowledge of Linux performance tuning, AI network topologies, and power/cooling for dense GPUs.
• Proven customer-facing technical leadership and ability to resolve complex infra issues.

Aufgaben

• Translate workload needs into GPU cluster architectures and capacity plans for deployments ranging from hundreds to thousands of GPUs.
• Recommend and implement infrastructure for training, inference, and HPC workloads, including orchestration and high‑speed interconnects.
• Configure and optimize production clusters using SLURM, Kubernetes, GPU drivers, and container runtimes.
• Diagnose issues across hardware, Linux, networking, drivers, and software; improve utilization and inter‑node performance.
• Build monitoring, alerting, automation, runbooks, and documentation; support critical deployments on call.
• Explain technical recommendations to customer engineers and executives; coordinate incident resolution with internal teams.

Kenntnisse

GPU clusters

HPC environments

SLURM

Kubernetes

InfiniBand

NVIDIA GPUs

CUDA

GPU drivers

Container runtimes

Ansible

Terraform

Python

Bash

Linux performance tuning

AI topology knowledge

Customer-facing leadership

Tools

NVIDIA DGX

Megatron‑LM

Jobbeschreibung

Role overview

Design and deploy GPU infrastructure for customers running large-scale AI training, inference, and high-performance computing workloads. This customer-facing role combines cluster architecture and capacity planning with hands-on implementation, performance tuning, and operational support across hardware and software.

Responsibilities

• Translate workload needs into GPU cluster architectures and capacity plans, including deployments ranging from hundreds to thousands of GPUs.


• Recommend and implement infrastructure for training, inference, and HPC workloads, including orchestration and high-speed interconnects.


• Configure and optimize production clusters using SLURM, Kubernetes, GPU drivers, and container runtimes.


• Diagnose issues across hardware, Linux, networking, drivers, and software; improve utilization and inter-node performance.


• Build monitoring, alerting, automation, runbooks, and operational documentation, and support critical deployments on call.


• Explain technical recommendations to customer engineers and executives, and coordinate incident resolution with internal teams.

Requirements

• At least three years of hands-on work with GPU clusters and HPC environments.


• Production experience with SLURM and Kubernetes, plus InfiniBand configuration and troubleshooting.


• Strong understanding of NVIDIA GPUs, CUDA, drivers, and GPU container runtimes.


• Experience with infrastructure automation such as Ansible or Terraform, and scripting or systems programming in Python and Bash.


• Knowledge of Linux performance tuning, AI network topologies, and power and cooling needs for dense GPU deployments.


• A record of customer-facing technical leadership and resolving complex infrastructure issues.

Nice to have

• Experience operating deployments with 1,000 or more GPUs, or relevant NVIDIA DGX, HGX, or SuperPOD certification.


• Familiarity with distributed training frameworks such as PyTorch FSDP, DeepSpeed, or Megatron-LM, and ML profiling.


• Experience with AMD MI300 or Intel Gaudi accelerators, or contributions to open-source HPC and AI infrastructure.

Benefits and work setup

• Remote work or an office option in San Francisco; visa sponsorship and relocation support are available.


• The stated cash range is $150,000–$300,000, with equity incentives, a professional development budget, and team off‑sites and conference attendance.