ML Platform Engineer — Distributed GPU Training on GKE

Egen; United States; A distancia; USD 159.000 - 219.000; Jornada completa

Ventajas ofrecidas por este puesto de trabajo

Comprehensive Health Insurance

Paid Vacation/PTO

401(k) Employer Match

Employee Referral Bonuses

Descripción de la vacante

Egen is seeking an engineer to own and optimize the GPU training platform embedded with a healthcare and life-sciences client. You will work on a large GKE-based H200/GB200 cluster, tuning workloads, benchmarking, and ensuring end-to-end performance.

You’ll collaborate with client scientists and other Egen engineers, owning your area outright. The role centers on the training stack, with strong emphasis on performance, observability, and scalable software practices.

Formación

• 5+ years of engineering experience with GPU training infrastructure.
• Strong Python software engineering with production-quality code and tests.
• GKE/Kubernetes on Google Cloud with GPU node pools and capacity management.
• Experience with PyTorch distributed training and Ray Train.
• Knowledge of GPU networking (RDMA, GPUDirect, NCCL) and benchmarking.

Responsabilidades

• Manage training platform on the GKE cluster and GPU node pools.
• Diagnose and fix platform failures at the root, not via restarts.
• Tune multi-node GPU networking for training throughput.
• Run, scale and debug multi-node distributed training with PyTorch and Ray Train.
• Profile jobs and remove bottlenecks across compute, memory, I/O, and interconnect.
• Instrument training with telemetry and maintain end-to-end tests.

Conocimientos

Python

Shell scripting

GPU training infra

Distributed training

Herramientas

Kubernetes on Google Cloud

Vertex AI

Ray

NCCL

GPUDirect

Descripción del empleo

Egen is seeking an engineer to own and optimize the GPU training platform embedded with a healthcare and life-sciences client. You will work on a large GKE-based H200/GB200 cluster, tuning workloads, benchmarking, and ensuring end-to-end performance.

You’ll collaborate with client scientists and other Egen engineers, owning your area outright. The role centers on the training stack, with strong emphasis on performance, observability, and scalable software practices.