AI Performance Engineer

Cornelis Networks, Inc.; Oregon; Hybrid; USD 120,000 - 190,000; Full time

Job summary

Cornelis Networks, Inc. is hiring an AI Performance Engineer to optimize training and multi-node inference across Omni-Path, Ethernet, and InfiniBand in GPU/CPU clusters. Collaborate with architecture, firmware, and software teams to turn lab results into field-proven improvements.

You will benchmark OpenAI/ML workloads, design efficient serving topologies, and visualize performance using Nsight, VTune, and Prometheus/Grafana. Remote work options available within the United States.

Qualifications

• BS in CS/EE/CE/Math or related (MS preferred) and 2-3+ years running AI/ML at cluster scale.
• Proven ability to set up, run, and analyze AI benchmarks; deep intuition for message passing, collectives, scaling efficiency, and bottleneck hunting for both training and inference.
• Hands-on with distributed training beyond single-GPU (DP/TP/PP, ZeRO, FSDP, sharded optimizers) and distributed inference architectures (replicated vs sharded, tensor/KV parallel, MoE).
• Practical experience across AI stacks & comms: PyTorch, DeepSpeed, Megatron-LM, RCCL/NCCL, vLLM, llama.cpp
• Comfortable with compilers(GCC/LLVM/Intel/OneAPI) and MPIstacks; Python + shell power user.
• Familiarity with network architectures (Omni-Path/OPA, InfiniBand, Ethernet/RDMA/RoCE) and Linuxsystems at the performance-tuning level, including NIC offloads, CQ moderation, pacing, ECN.
• Excellent written and verbal communication—turn measurements intocustomer persuasion.

Responsibilities

• Measure & optimize end-to-end performance for distributed AI workloads (training + multi-node inference) across CPU & GPU clusters and fabrics (Omni-Path, Ethernet, InfiniBand).
• Benchmark, characterize, and tune workloads and benchmarks (e.g., Llama, Qwen) on current/future compute, storage, and network hardware; frameworks include Megatron-LM, TorchTitan, Composer, vLLM, llama.cpp.
• Design and optimize distributed serving topologies (sharded/replicated, tensor/pipe parallel, MoE), batching, KV-cache sharding/offload (CPU/NVMe) & prefix caching.
• Optimize training & inferencing: Validate RDMA/GPUDirect RDMA, congestion control, and collective/point-to-point tradeoffs during inference.
• Build proof points comparing Cornelis Omni-Path to competing interconnects; translate data into narratives for sales/marketing and customers.
• Instrument and visualize performance (Nsight Systems, VTune, perf, app timers) plus Prometheus/Grafana visuals.
• Evangelize best practices through briefs, READMEs, blogs, articles, and peer-reviewed publications.

Skills

AI benchmarks

Distributed training

MPI

PyTorch

DeepSpeed

Megatron-LM

NCCL

Python

Education

Bachelor's degree in CS/EE/CE/Math

MS preferred

Tools

Nsight Systems

VTune

Prometheus Grafana

Nsight Compute

rocprof/roctracer

GCC/LLVM/OneAPI

Job description

Cornelis Networks delivers the world's highest performance scale-out networking solutions for AI and HPC datacenters. Our differentiated architecture integrates hardware, software, and system technologies to maximize the efficiency of GPU, CPU, and accelerator-based clusters at any scale. We're committed to innovation, performance, and scalability—helping customers push the boundaries of what's possible.We are a fast-growing, forward-thinking team of architects, engineers, and business professionals with a proven track record of building successful products and companies. As a global organization, our team spans multiple U.S. states and six countries, and we continue to expand with exceptional talent in onsite, hybrid, and fully remote roles.We're seeking an AI Performance Engineer. Join us to optimize training and multi-node inference across next-gen networking silicon and systems—adapters, switches, and the software stack that ties it all together. You'll partner with architecture, firmware, software, and lighthouse customers to turn lab results into field-proven wins.

Key Responsibilities

• Measure & optimize end-to-end performance for distributed AI workloads (training + multi-node inference) across CPU & GPU clusters and diverse fabrics (Omni-Path, Ethernet, InfiniBand).
• Benchmark, characterize, and tune open-source & industry workloads and benchmarks (e.g., Llama, Qwen) on current and future compute, storage, and network hardware, including frameworks such as Megatron-LM, TorchTitan, Composer, vLLM, llama.cpp, etc.
• Design and optimize distributed serving topologies (sharded/replicated, tensor/pipe parallel, MoE expert placement), continuous/adaptive batching, KV-cache sharding/offload (CPU/NVMe) & prefix caching.
• Optimize training & inferencing: ValidateRDMA/GPUDirect RDMA, congestion control, and collective/point-to-point tradeoffs during inference.
• Build crisp proof points that compare Cornelis Omni-Path to competing interconnects; translate data into narratives for sales/marketing and existing customers.
• Instrument and visualize performance (Nsight Systems, VTune, perf, app timers) plus database visualization (Prometheus/Grafana).
• Evangelize best practices through briefs, READMEs, blogs, articles, and peer-reviewed publications

Minimum Qualification

• BS in CS/EE/CE/Math or related (MS preferred) and 2-3+ years running AI/ML at cluster scale.
• Proven ability to set up, run, and analyze AI benchmarks; deep intuition for message passing, collectives, scaling efficiency, and bottleneck hunting for both training and inference.
• Hands-on with distributed training beyond single-GPU (DP/TP/PP, ZeRO, FSDP, sharded optimizers) and distributed inference architectures (replicated vs sharded, tensor/KV parallel, MoE).
• Practical experience across AI stacks & comms:PyTorch, DeepSpeed, Megatron-LM, RCCL/NCCL, vLLM, llama.cpp
• Comfortable with compilers(GCC/LLVM/Intel/OneAPI) and MPIstacks; Python + shell power user.
• Familiarity with network architectures (Omni-Path/OPA, InfiniBand, Ethernet/RDMA/RoCE) and Linuxsystems at the performance-tuning level, including NIC offloads, CQ moderation, pacing, ECN.
• Excellent written and verbal communication—turn measurements intocustomer persuasion.

Preferred Qualifications

• Cluster scheduler experience (e.g. SLURM, PBS).
• Hands-on profiling & tracing of GPU/comm paths (Nsight Systems, Nsight Compute, ROCm tools/rocprof/roctracer, VTune, perf, etc.).
• Background in HPC performance engineering or storage (BeeGFS, Lustre, NVMe-oF).

LocationRemote:

This role fully supports remote work for employees residing within the United States.Cornelis is a company that is strongly led by our core values and believes in the responsibility of our people, which is why we fully support remote employees. When safe, we encourage travel for partner and client meetings, conferences, and company in-person collaboration and meetings. Expect travel to be less than 10%.

Cornelis Networks is an equal opportunity employer, and all qualified applicants will receive consideration for employment without regard to race, color, religion, sex, sexual orientation, gender identity or expression, pregnancy, age, national origin, disability status, genetic information, protected veteran status, or any other characteristic protected by law.