GPU Cluster Architect: Scalable AI Platform

Sciforium; San Francisco (CA); On-site; USD 190,000 - 270,000; Full time

Benefits offered by this job

Medical insurance

401k plan

Daily meals/snacks

Flexible time off

Competitive salary and equity

Job summary

Sciforium is seeking a GPU Cluster Engineer to own the software stack of GPU clusters—from kernel tuning to ML framework integration. You will define production-ready nodes, automate bring-up, and maintain fleet health while supporting foundation model training and model serving teams.

You will automate image creation, validation suites, and fleet upgrades, leveraging Kubernetes, Slurm, and container tooling to ensure high performance and reliability across the cluster.

Qualifications

• 5+ years in systems/infrastructure engineering with GPU cluster/ML infra experience.
• Bachelor's or Master's in CS/EE or related field.
• Deep Linux internals: kernel modules, DKMS, NUMA, cgroups, system performance tuning.
• Hands-on experience with NVIDIA CUDA and/or ROCm driver stacks on modern accelerators.
• Production Kubernetes experience with GPU workloads; knowledge of HPC schedulers (Slurm/Run:AI).
• Configuration management with Ansible or SaltStack; Git-based workflows.
• Image provisioning/tooling (Packer, MaaS, Foreman, Terraform).
• Experience with distributed filesystems (Lustre, GPFS) and checkpoint I/O optimization.
• Python and Bash for automation; NCCL/RCCL and PyTorch/JAX runtime familiarity.

Responsibilities

• Own node software definition from OS bring-up to production-ready GPU nodes.
• Build automated acceptance suites and validation for fleet readiness.
• Maintain fleet through kernel/driver/toolkit upgrades with minimal disruption.
• Automate health checks, drift detection, and repair handoffs.
• Manage node/configuration via IaC (Ansible/SaltStack) with CI validation and canaries.
• Develop tooling for cluster operations, health reporting, and workflows.
• Operate GPU-enabled Kubernetes and Slurm for serving and training workloads.
• Maintain CUDA/ROCm stacks, container tooling, and reproducible images.

Skills

GPU cluster

Linux internals

Python scripting

Kubernetes

NVIDIA CUDA

Ansible

System performance

Education

Bachelor's or Master's in CS/EE

Tools

NVIDIA Driver Toolkit

ROCm

Packer

Terraform

Job description

Sciforium is seeking a GPU Cluster Engineer to own the software stack of GPU clusters—from kernel tuning to ML framework integration. You will define production-ready nodes, automate bring-up, and maintain fleet health while supporting foundation model training and model serving teams.

You will automate image creation, validation suites, and fleet upgrades, leveraging Kubernetes, Slurm, and container tooling to ensure high performance and reliability across the cluster.