Senior Platform Engineer: GPU Compute & Orchestration

Perplexity; Sacramento (CA); On-site; USD 200,000 - 300,000; Full time

Job summary

Perplexity is seeking an experienced platform engineer to own and evolve a self-serve compute platform for training and inference workloads on a GPU fleet. You will design systems to deploy and manage GPU resources across providers and build the Kubernetes-based orchestration that powers both long-running training jobs and production inference services.

The role requires deep Kubernetes expertise, experience with GPU hardware and networking, and the ability to coordinate across clouds.

Qualifications

• Deep Kubernetes experience with operators, CRDs, and multi-cluster federation.
• Experience managing GPU clusters at scale (NVIDIA hardware, CUDA, fast network).
• Orchestrated compute across multiple clouds (e.g., AWS, GCP) with understanding of cloud differences.
• Strong distributed-systems fundamentals: scheduling, allocation, and fault tolerance.

Responsibilities

• Build a self-serve compute platform for training and inference workloads.
• Design systems to launch training jobs and operate inference services without GPU provisioning overhead.
• Operate the GPU fleet and own provisioning, lifecycle, and capacity across providers.
• Develop scheduling and placement logic to maximize capacity and match workloads to hardware under constraints.
• Make failure boring by building fault tolerance, autoscaling, and observability across the fleet.
• Own Kubernetes orchestration for GPUs and write operators/CRDs; manage multiple clusters across providers.
• Set technical direction and collaborate with inference and cloud infra teams to shape platform architecture.

Skills

Kubernetes

GPU clusters

Multi-cloud

Distributed systems

Go/Rust/C++

End-to-end ownership

Tools

Kubernetes Operators

CRDs

Multi-cluster federation

NVIDIA CUDA

InfiniBand

RoCE

Job description

Perplexity is seeking an experienced platform engineer to own and evolve a self-serve compute platform for training and inference workloads on a GPU fleet. You will design systems to deploy and manage GPU resources across providers and build the Kubernetes-based orchestration that powers both long-running training jobs and production inference services.

The role requires deep Kubernetes expertise, experience with GPU hardware and networking, and the ability to coordinate across clouds.