ML Platform Engineer, Distributed Training

The ML Platform Engineer will support a large scale GPU training platform on Google Cloud for foundation model development in healthcare and life sciences. Responsibilities include tuning GKE GPU scheduling, multi node networking, RDMA, GPUDirect, and NCCL to improve distributed PyTorch and Ray Train performance. The role will profile and debug training workloads, build benchmarks and evaluation harnesses, and implement telemetry and end to end testing. Success requires strong Python engineering, ownership of platform improvements, and close collaboration with infrastructure teams and research scientists.