DevOps & Infrastructure Engineer (HPC/GPU)

You will own the infrastructure that powers our R&D and helps our customers deploy our technology on-premise. You will move beyond standard cloud DevOps into the world of High-Performance Computing (HPC).

• Think: Design a robust CI/CD strategy that handles cross-platform compilation (Windows/Linux) and execution on specific hardware targets (NVIDIA A100, AMD MI250, Consumer GPUs). Architect solution templates for our customers who need to deploy Hybridizer-generated binaries on their own private clouds.

• Implement:

  • Set up and maintain Kubernetes clusters (both on-premise and cloud) with GPU Passthrough and Multi-Instance GPU (MIG) configurations.

  • Develop GitHub Actions pipelines that seamlessly dispatch heavy test suites to self-hosted runners equipped with specific GPU accelerators.

  • Configure DockerHub registries and secure container lifecycles for our compiler images.

• Build:

  • Hardware Tuning: Assemble and fine-tune physical servers. This includes managing PCIe topology, cooling profiles, and power constraints to ensure consistent benchmarking results.

  • Driver Ecosystem: Manage the complex matrix of NVIDIA drivers, CUDA toolkits, and ROCm versions across our fleet, ensuring compatibility with our compiler’s output.

You are a DevOps engineer who loves hardware. You understand that "the cloud" is just someone else's computer, and sometimes you need to manage that computer yourself.

• Core DevOps: Strong mastery of Docker and Kubernetes. You know how to write custom Helm charts and manage stateful sets.

• GPU Infrastructure: You have hands-on experience with NVIDIA Container Toolkit or ROCm integration in containers. You understand concepts like PCIe passthrough, IOMMU groups, and GPU orchestration.

• CI/CD Automation: Expert in GitHub Actions. You can write complex workflows with matrix strategies and self-hosted runners.

• System Administration: You are comfortable with Linux kernel tuning, driver installation (dkms), and diagnosing hardware bottlenecks.

• Customer Facing: You have the communication skills to assist clients. You can explain how to expose a GPU to a Docker container to a sysadmin who might not be an expert in HPC.

• Adaptability: You are ready to work with a mix of consumer and data-center grade hardware (e.g., configuring a server with 4x RTX 5090s or managing a DGX station).