Remote GPU & Cloud Infrastructure Engineer for AI

Innomium; Northern (KY); Hybrid; USD 150,000 - 210,000; Full time

Benefits offered by this job

Remote-first policy

Annual learning budget

Company-provided equipment

Internet/home-office stipend

Paid time off

Equity/bonuses where applicable

Retirement contributions

Job summary

Innomium is seeking a GPU and Cloud Infrastructure Engineer to build the compute foundation for model teams, covering training, evaluation, fine-tuning, and inference.

You will design GPU workloads, containerized environments, storage and networking paths, and automation to reproduce experiments and releases across programs, while collaborating with researchers to optimize utilization, memory, and cost.

Qualifications

• Experience with cloud infrastructure for GPU workloads.
• Strong Linux, containers, networking, storage.
• IaC fundamentals across cloud platforms.
• Ability to diagnose failures across infra layers and document work.

Responsibilities

• Build and operate GPU training, evaluation, fine-tuning, and inference environments.
• Automate provisioning, container images, dependencies, secrets, networking, and storage.
• Profile utilization, memory, throughput, queue behavior, data transfer, and cost.
• Design model-serving and batch-execution paths with observability and recovery.
• Collaborate on CUDA, PyTorch, kernel, and framework compatibility.
• Create runbooks, capacity models, security controls, and reproducible docs.
• Improve developer workflows for launching and debugging GPU jobs.
• Surface cost and performance trade-offs to stakeholders.

Skills

GPU workloads

Linux

Containers

Networking

Storage

Infrastructure as code

Cloud platforms

Container orchestration

NVIDIA drivers/CUDA

PyTorch

Diagnostics

Tools

Kubernetes GPU scheduling

Slurm

Ray

Job description

Innomium is seeking a GPU and Cloud Infrastructure Engineer to build the compute foundation for model teams, covering training, evaluation, fine-tuning, and inference.

You will design GPU workloads, containerized environments, storage and networking paths, and automation to reproduce experiments and releases across programs, while collaborating with researchers to optimize utilization, memory, and cost.