GPU Infrastructure Engineer for ML Systems (Hybrid)

H Company; Paris; Hybrid; GBP 90,000 - 140,000; Full time

Benefits offered by this job

Competitive package

Job summary

H Company is building a globally distributed GPU compute platform. You’ll own the GPU compute that powers model training, evaluation and serving systems, and you’ll optimize scheduling, node health, and capacity across multiple regions in a hybrid setup.

You will work with Kubernetes, cloud providers and vendor teams to keep production reliable and to improve performance of training and serving stacks. Strong Python skills and GPU experience are essential.

Qualifications

• Experience running GPU workloads on Kubernetes clusters.
• Ability to design GPU clusters for training and online RL.
• Strong debugging skills at OS/network/storage layers.

Responsibilities

• Run Kubernetes clusters for GPU training and inference.
• Improve GPU scheduling and placement efficiency.
• Detect, drain and repair unhealthy GPU nodes.
• Manage capacity with cloud providers.
• Debug failures across GPUs, Linux, storage and networking.

Skills

GPU workloads on Kubernetes

Python

OS debugging

Tools

Slurm

Kubernetes scheduling

Kueue

Karpenter

SkyPilot

AWS EKS

SageMaker HyperPod

NCCL

Open MPI

Terraform

InfiniBand

EFA

Job description

H Company is building a globally distributed GPU compute platform. You’ll own the GPU compute that powers model training, evaluation and serving systems, and you’ll optimize scheduling, node health, and capacity across multiple regions in a hybrid setup.

You will work with Kubernetes, cloud providers and vendor teams to keep production reliable and to improve performance of training and serving stacks. Strong Python skills and GPU experience are essential.