Senior GPU Infrastructure Engineer

Career Movement; San Francisco (CA); On-site; USD 180,000 - 240,000; Full time

Job summary

Career Movement is seeking a senior engineer to power the compute behind its research and products in the Bay Area. You will write production code to schedule, serve and repair fleets of accelerators, taking ownership from day one and delivering visible impact on critical runs.

You will build and operate GPU clusters across public clouds and owned data centers, design schedulers, stand up model serving and distributed training environments, and report on spend with a focus on reliability and

Qualifications

• 5+ years building production distributed systems or ML infrastructure.
• Deep Kubernetes experience, with exposure to Slurm or similar schedulers.
• Proven experience running large GPU workloads on a modern training or serving stack.
• Willingness to carry the pager for systems that other teams depend on.

Responsibilities

• Build and operate GPU clusters across public cloud, specialty GPU providers and owned data centers.
• Design schedulers and automation that increase utilization and recover from hardware failures without manual intervention.
• Stand up fast model serving and distributed training environments.
• Report on where compute spend goes and where it can be reduced.

Skills

Distributed systems

Kubernetes

GPU workloads

Pager duty

Tools

Slurm

Job description

A profitable, high growth technology company in the Bay Area is expanding the team that powers the compute behind its most important research and products. As a senior engineer, you will write production code that decides how large fleets of accelerators are scheduled, served and repaired. You will become the engineer researchers call when a critical run has to go right, with real ownership and visible impact from day one.

What you will do:

• Build and operate GPU clusters across public cloud, specialty GPU providers and owned data centers
• Design schedulers and automation that increase utilization and recover from hardware failures without manual intervention
• Stand up fast model serving and distributed training environments
• Report on where compute spend goes and where it can be reduced

What you bring:

• 5+ years engineering production distributed systems or ML infrastructure
• Deep Kubernetes experience, with exposure to Slurm or similar schedulers
• Proven experience running large GPU workloads on a modern training or serving stack
• Willingness to carry the pager for systems that other teams depend on