ML Platform Engineer

Job Summary

To support a healthcare and life-sciences AI client, the full-time ML Platform Engineer will manage GPU training platforms, scale and debug multi-node distributed training, and design performance benchmarks while working remotely.

Key responsibilities

• Collaborate with the client's infrastructure team to optimize the GKE training platform for workload performance
• Run and debug multi-node distributed training jobs using PyTorch and Ray Train, profiling and resolving bottlenecks
• Design and implement performance benchmarks and evaluation harnesses to assess the training stack's efficiency

Required qualifications

• 5+ years of engineering experience with a focus on GPU training infrastructure and large-scale training
• Strong proficiency in Python and experience with production-quality code and testing
• Hands-on experience with Kubernetes on Google Cloud, particularly with GPU scheduling and job queueing
• Experience running and debugging multi-node training jobs with Distributed PyTorch and Ray Train
• Knowledge of RDMA and GPU networking, including GPUDirect and NCCL tuning