Remote Senior AI & HPC Systems Engineer

cornelisnetworks; Austin (TX); On-site; USD 180,000 - 240,000; Full time

Job summary

Cornelis Networks is seeking a highly experienced Senior Software Engineer to design, develop, and upstream-enable its AI communication middleware for distributed AI workloads. You will optimize collectives (NCCL/RCCL) over Cornelis networks interconnects and collaborate across kernel/driver, switch, and systems teams.

This remote US position offers equity, compensation, and a generous benefits package. The role requires 8+ years in high-performance C/C++ on Linux, strong GPU communication stack

Qualifications

• 8+ years in high-performance systems programming on Linux.
• Strong experience with GPU communication stacks (CUDA/ROCm, NCCL/RCCL).
• Ability to optimize distributed training performance with profiling/tracing.
• Understanding of collective communication concepts and topology.

Responsibilities

• Design and implement performance-critical features for AI communication middleware.
• Optimize distributed training performance across multi-node, multi-GPU setups.
• Improve GPU communication paths (GPU-direct transfers, IPC, CPU/GPU sync).
• Profile workloads and identify bottlenecks across software/hardware stack.
• Prototype and validate Ultra Ethernet capabilities for AI collectives.
• Collaborate with kernel/driver, switch, and systems teams, and review code.

Skills

C/C++ Linux

CUDA ROCm NCCL RCCL

Profiling & tracing

Collective communication

Production-grade code

Open-source contributions

Tools

libfabric/OFI

UCX

RDMA

Job description

Cornelis Networks is seeking a highly experienced Senior Software Engineer to design, develop, and upstream-enable its AI communication middleware for distributed AI workloads. You will optimize collectives (NCCL/RCCL) over Cornelis networks interconnects and collaborate across kernel/driver, switch, and systems teams.

This remote US position offers equity, compensation, and a generous benefits package. The role requires 8+ years in high-performance C/C++ on Linux, strong GPU communication stack