System Administrator

OXMIQ designs GPU and AI silicon for large-scale model inference and training and is developing an infrastructure and AI service orchestration platform that runs on heterogeneous accelerator hardware.

The System Administrator owns the day-to-day health of OXMIQ's office, lab, and supporting IT environments — the infrastructure through which our engineers, executives, and customer-facing teams get their work done. The role is responsible for the end-to-end operational path: how a system is imaged, deployed, networked, monitored, supported, and retired across our Campbell headquarters and Fremont colocation footprint.

The System Administrator must also have a working understanding of the AI workloads that run on top of this infrastructure — model serving, AI agents, GPU runtimes — and will collaborate with the platform and engineering teams to deploy, operate, and troubleshoot these services. Background and hands-on experience with these layers is expected; ownership of their delivery is not.

The role is hands‑on. The System Administrator builds and maintains imaging pipelines, configures network and conference room infrastructure, onboards new employees, deploys AI services on internal hardware, and serves as the first line of IT support for the engineering organization.

• Own the bare-metal imaging and provisioning pipeline for the office and lab fleet, including FOG Server (Windows/Linux workstations) and MAAS (Ubuntu/RHEL servers), golden image management, PXE boot infrastructure, and post-deployment automation.
• Configure, monitor, and maintain office and colocation network infrastructure, including Forti Gate firewalls, switches, VLANs, Wi‑Fi, DNS, DHCP, VPN, and site-to-site tunnels (Tailscale, devtunnel, IPsec).
• Patch, monitor, and maintain Windows, Ubuntu, and RHEL systems across the office and lab; manage backups, TrueNAS storage, hardware inventory, lifecycle replacement, and disaster recovery procedures.
• Provision laptops, accounts, and access for new hires end-to-end; maintain onboarding runbooks and standard hardware/software configurations; conduct day‑one IT orientation and ongoing user support.
• Configure and maintain conference room systems (displays, cameras, microphones, room scheduling) and support hybrid meetings across Teams, Zoom, and Google Meet for executive and customer‑facing use.
• Deploy, configure, and maintain AI agents and AI gateway services (e.g., Open Claw) across internal infrastructure; integrate with internal identity, networking, and observability stacks.
• Stand up and operate self-hosted LLM endpoints (vLLM, SGLang, llama.cpp, or comparable) on the GPU fleet; pull, quantize, and validate models from Hugging Face and other sources; manage model storage and versioning.
• Support engineers in benchmarking and evaluating models across NVIDIA, AMD, Intel, and Tenstorrent hardware; monitor GPU utilization, inference latency, and endpoint health; troubleshoot device pairing, drivers, and runtime issues.
• Serve as the first line of IT support for the engineering team on hardware, software, and access issues; document procedures, write runbooks, and contribute to the internal knowledge base.

• 3+ years of system administration experience in a mixed Windows/Linux environment, with a track record of owning end-to-end IT operations.
• Hands‑on experience with bare‑metal imaging and provisioning systems — FOG Server, MAAS, or comparable (Clonezilla, SCCM/MECM, Foreman) — at the level of operating them in production, customizing images, and integrating with PXE/DHCP infrastructure.
• Strong networking fundamentals: VLANs, routing, firewalls, DNS, DHCP, VPN, and the operational realities of running a multi‑site network across office and colocation facilities.
• Proficiency with at least one configuration management tool (Ansible, Puppet, Chef) and comfort scripting in Power Shell and Bash.
• Hands‑on experience supporting conference room AV and end‑user IT in a professional environment, including Teams, Zoom, and Google Meet integrations.
• Working knowledge of cloud administration in at least one of Azure, GCP, or AWS, including identity federation…