Member of Technical Staff (Software Engineer, GPU Cluster Infrastructure)

The role

You will build a self-serve compute platform to manage a large GPU fleet across multiple cloud providers. You will own the infrastructure to support training and inference workloads.

What you will do

• Build a self-serve compute platform.
• Operate the GPU fleet including provisioning, lifecycle management, reliability, and capacity integration.
• Build scheduling and placement logic to solve for GPU scarcity.
• Support long-running distributed training jobs and production inference services.
• Own Kubernetes for GPU orchestration including writing operators and CRDs.
• Build fault tolerance, autoscaling, and observability.
• Set technical direction by partnering with inference and cloud infrastructure engineers.

Required qualifications

• Deep Kubernetes experience including custom operators, CRDs, and multi-cluster federation.
• Managed GPU clusters at scale using NVIDIA hardware, CUDA, and networking like InfiniBand or RoCE.
• Orchestrated compute across multiple clouds such as CoreWeave, AWS, or GCP.
• Strong distributed systems fundamentals in scheduling, resource allocation, and fault tolerance.
• Write infrastructure and systems-level code in Go, Rust or C++.
• Supported both long-running training jobs and high-availability inference services.
• Own problems end-to-end.

Nice to have

• Inference serving stacks: vLLM, SGLang, or TensorRT-LLM.
• Slurm or other HPC schedulers.
• GPU kernel work in CUDA or Triton.
• High-speed interconnects: InfiniBand, RoCE, or RDMA in production.
• Observability for ML workloads: Prometheus, Grafana, or Weights & Biases.

About the company

Perplexity serves hundreds of millions of queries a month using a large GPU fleet spread across several cloud providers.