AI Performance Engineer

Location: Remote within the United States; based in the San Jose, California performance department. Travel for partner and client meetings, conferences, and in-person collaboration is expected to be less than 10%.

Company

hirify.global develops scale-out networking solutions for AI and HPC datacenters by integrating hardware, software, and system technologies for GPU, CPU, and accelerator clusters.

What you will do

• Measure and optimize end-to-end performance for distributed AI training and multi-node inference across CPU/GPU clusters and Omni-Path, Ethernet, and InfiniBand fabrics.
• Benchmark and tune AI workloads, frameworks, compute, storage, and networking hardware, including Llama, Qwen, Megatron-LM, TorchTitan, Composer, vLLM, and llama.cpp.
• Design and optimize distributed serving topologies, parallelism strategies, MoE expert placement, continuous and adaptive batching, KV-cache sharding/offload, and prefix caching.
• Validate RDMA, GPUDirect RDMA, congestion control, and collective versus point-to-point tradeoffs; investigate scaling efficiency and bottlenecks.
• Build performance comparisons between Cornelis Omni-Path and competing interconnects, translating measurements into customer, sales, and marketing narratives.
• Instrument and visualize performance using Nsight Systems, VTune, perf, application timers, Prometheus, and Grafana, and share best practices through technical documentation and publications.

Requirements

• Bachelor’s degree in Computer Science, Electrical Engineering, Computer Engineering, Mathematics, or a related field; a master’s degree is preferred.
• 2–3+ years of experience running AI/ML workloads at cluster scale.
• Proven ability to set up, run, and analyze AI benchmarks, with deep understanding of message passing, collectives, scaling efficiency, and bottleneck analysis for training and inference.
• Hands-on distributed training experience beyond a single GPU, including DP, TP, PP, ZeRO, FSDP, and sharded optimizers, plus distributed inference architectures such as replicated and sharded serving, tensor/KV parallelism, and MoE.
• Practical experience with PyTorch, DeepSpeed, Megatron-LM, RCCL/NCCL, vLLM, and llama.cpp.
• Experience with GCC, LLVM, Intel, or OneAPI compilers; MPI stacks; and Python and shell scripting.
• Familiarity with Omni-Path/OPA, InfiniBand, Ethernet, RDMA/RoCE, and Linux performance tuning, including NIC offloads, CQ moderation, pacing, and ECN.
• Excellent written and verbal communication skills for turning measurements into customer persuasion.

Nice to have

• Cluster scheduler experience with SLURM, PBS, or similar tools.
• Hands-on GPU and communication-path profiling and tracing with Nsight Systems, Nsight Compute, ROCm tools, rocprof, roctracer, VTune, perf, or similar tools.
• Background in HPC performance engineering or storage technologies such as BeeGFS, Lustre, or NVMe-oF.

Culture & Benefits

• Remote work is fully supported for employees residing within the United States.
• Occasional travel is supported for partner and client meetings, conferences, and in-person collaboration, with travel expected to be less than 10%.
• Work with architecture, firmware, software, and lighthouse customer teams on field-proven performance improvements.
• Join a global organization spanning multiple U.S. states and six countries, with onsite, hybrid, and fully remote roles.