Sfcompute The HPC Data Center Ops Engineer at Sfcompute is responsible for hands-on operation and maintenance of GPU clusters and data center hardware. This includes racking, commissioning, and replacing failed components.
Responsibilities
Our clusters don’t run themselves. Someone has to pull the failed tray, chase the bad optic that’s flapping a link at 3am, and make sure the spare was already on the shelf before anyone needed it. That’s this role, and it’s as hands-on as it gets.
From racking and commissioning systems to replacing failed components, managing fiber, transceivers, liquid cooling, and maintaining accurate asset records. You’ll also manage procurement, including BOMs, vendor quotes, lead times, spares, warranties, and RMAs.
Around 80% of the role focuses on data center operations and procurement, with 20% supporting office IT. Regular visits to local facilities and occasional travel to remote sites are required.
You’ll work with cutting-edge infrastructure, including HGX/DGX-class GPU systems, InfiniBand and 400G+ networks, high-density power, and liquid cooling. As part of a small team, you’ll also help establish standards, develop runbooks, and mentor future team members.
Qualification
You’ve replaced server FRUs and rackedYou’ve bought hardwareYou’re comfortable with TCP/IPSome Nice To HavesExperience in HPCExperience with liquid cooling — CDUsNetBox or similar DCIM
Required
You know fiber — high-density trunks and breakouts, end-face inspection and cleaning, transceiver swaps — and you can tell a dirty connector from a dying optic
You’ve replaced server FRUs and racked, cabled, and commissioned equipment, and you work safely in and around energized racks
You’ve bought hardware, not just installed it: BOMs, vendor quotes, POs, receiving, warranty, and RMAs are familiar territory
You keep inventory and asset records that would survive an audit, across more than one site
You’re comfortable with TCP/IP, DHCP, DNS, VLANs, out-of-band management (IPMI, iDRAC, iLO, Redfish), and the Linux command line
You’re open to occasional travel to remote sites and to after-hours maintenance windows when the work requires it
Some Nice To Haves
Experience in HPC, GPU, or other high-density compute environments, including HGX- or DGX-class platforms and rack-scale designs
Experience with InfiniBand or high-rate Ethernet at 400G and above
Experience with liquid cooling — CDUs, manifolds, and direct-to-chip cold plates
Colocation experience and familiarity with the tenant/operator relationship
NetBox or similar DCIM, and procurement or ERP systems
