Compute Infrastructure Engineer for Frontier AI

openai; General Trias; On-site; PHP 11,243,000 - 18,738,000; Full time

Job summary

OpenAI is seeking engineers to build the compute platform behind its research and products. You may work near hardware or near users, on CaaS and agent infrastructure, or on the control and data planes in between.

You could help bring new supercomputing capacity online, optimize training workloads, improve NCCL and collective communication, and design abstractions that make heterogeneous clusters feel like one coherent platform.

Qualifications

• Strong software engineering skills with production infrastructure experience.
• Experience in distributed systems, HPC, GPU infrastructure, or related areas.
• Ability to debug complex system behavior across software, hardware, networking, and workload layers.

Responsibilities

• Build and optimize reliable system software for large-scale compute systems.
• Design and operate infrastructure across accelerators, CPUs, NICs, networking, storage, data centers, cluster orchestration, scheduling, and fleet health.
• Profile, benchmark, and optimize training workloads and scheduling bottlenecks across compute, memory, storage, networking, NCCL.
• Create hardware-aware automation for provisioning, firmware and driver upgrades, incident response, and day-to-day operations.
• Build CaaS, agent infrastructure, profiling, observability, benchmarking, and platform tools for researchers, engineers, and operators.

Skills

Distributed systems

Operating systems

Networking protocols

High-performance computing

GPU infrastructure

CaaS

Agent infrastructure

Scheduling

Reliability engineering

Developer platforms

Instrumentation & observability

Tools

Kubernetes

NCCL

RDMA

GPU tooling

Observability tooling

Job description

OpenAI is seeking engineers to build the compute platform behind its research and products. You may work near hardware or near users, on CaaS and agent infrastructure, or on the control and data planes in between.

You could help bring new supercomputing capacity online, optimize training workloads, improve NCCL and collective communication, and design abstractions that make heterogeneous clusters feel like one coherent platform.