Production Engineer, Network Engineering

Metric DCX; Austin (TX); On-site; USD 120,000 - 150,000; Full time

Job summary

Metric DCX seeks a Production Engineer, Network Engineering to own pager-on-call duties, diagnose outages, and build tooling that makes repairs repeatable for fast-scaling AI infrastructure across large datacenters.

You will write Python and Go tooling to automate link tests, remote command execution, and repair visualization, while maintaining real-time monitoring and health dashboards for ongoing operations. Strong scripting and network diagnostics are essential in this multi-site role.

Qualifications

• Experience carrying a pager for production networks.
• Proficiency in Python or Go scripting for automation.
• Hands-on with network diagnostics, optics, and monitoring (gNMI/NETCONF).

Responsibilities

• Take the on-call pager and diagnose the fault end-to-end, executing fixes or driving RMAs.
• Write Python and Go tooling to automate network tasks and repairs.
• Automate the repair pipeline from fault detection through RMA initiation and ticket integration.
• Maintain real-time monitoring and alerting to reflect fleet health across sites.
• Validate new sites and hardware with qualification tests before carrying traffic.

Skills

Pager on-call

Python

Go

Network diagnostics

gNMI

gRPC

NETCONF

SSH

Tools

CLI

Spine-Leaf

Transceiver testing

Remote command execution

Job description

We are seeking a Production Engineer, Network Engineering to carry the pager, repair end to end, and build the tooling that makes that repair repeatable for a rapidly scaling AI infrastructure company building large, high-density datacenter sites and the networks that connect them.

About the company: A fast-growing team designing, building, and operating power and data infrastructure for large-scale AI compute. The work is greenfield and multi-site, at a scale where manual operations do not hold up in production.

What you will do:

• Carry the on-call pager and run repair end to end: diagnose the fault, execute the fix or drive the RMA, and return the link to service across DC fabric, edge, and host-to-network
• Write Python and Go tooling that replaces manual diagnosis, including link tests, remote command execution across the fleet, and repair visualization that shows what is broken and why
• Automate the repair pipeline from fault detection through RMA initiation, ticket integration, transceiver and optics tracking, and return to service, so a failure at any site follows the same path
• Maintain the real-time monitoring and alerting that gives every on-call engineer a true picture of network health, and keep it accurate as new sites and hardware roll in
• Validate new sites and hardware into production with the qualification tests that confirm a network is healthy before it carries traffic

What we are looking for:

• You have carried a pager for a production network and can walk into an outage, find the fault, fix it or drive the RMA, and write the postmortem without someone walking you through it
• You have written Python or Go that replaced a manual network task, from link diagnostics to config pushes to fleet-wide command execution
• You can tell a transceiver fault, a bad route, and a power event apart from the signals alone
• Hands-on with link diagnostics, optics, and network monitoring (gNMI, gRPC, NETCONF, SONiC), and comfortable at the CLI on switches and routers across a fleet
• You treat toil as a bug. If a repair step means SSHing into ten boxes by hand, you script it once
• You get competent in an unfamiliar part of the stack fast, and you document it so the next on call does not start from zero
• Bonus points: RMA and repair lifecycle automation, large-scale datacenter fabric (BGP, ECMP, spine-leaf), out-of-band network management, and daily use of AI coding tools

Why this role:

• Own repair from first alert to return to service, not just the ticket
• Build the tooling that turns a network fault from a mystery into a closed loop
• Work on mission-critical infrastructure where a new site does not wait for process to catch up