Lead Inference Performance Architect

Neurophos, Inc.; Sunnyvale, Northern (TX, KY); Hybrid; USD 150,000 - 210,000; Full time

Benefits offered by this job

Health benefits

Unlimited PTO

401(k) matching

Stock options

Voluntary benefits

Personalized benefits

Job summary

Neurophos, Inc. is seeking a performance engineer to own benchmarking and energy metrics for the T100 optical inference accelerator.

You will define measurement methods, generate end-to-end results, and ensure reproducibility across models, fidelities, and hardware generations. The role requires 5+ years in GPU performance, with hands-on experience in roofline and analytical modeling, and the ability to work with cloud GPUs and ML stacks.

Qualifications

• BS or MS in Computer Engineering, Electrical Engineering, Computer Science or equivalent practical experience
• 5+ years of experience in GPU performance engineering, accelerator benchmarking, HPC performance measurement, or ML systems measurement
• Track record of building or operating benchmark harnesses that produced measured results on real GPUs or accelerators
• Hands-on experience with roofline analysis, limiter analysis, or analytical performance modeling
• Proficiency in Python for harnesses, parsing, and plots, and comfort working in Linux
• Cloud GPU operations on AWS, GCP, or Azure, including containers, instance types, drivers, quotas, and cost

Responsibilities

• Own the performance and energy metrics that architecture, product, and leadership rely on, across modeling fidelities and hardware
• Produce numbers for the same workloads across fidelities: roofline, limiter models, architecture models, RTL simulation, and measured hardware
• Keep workload definitions constant across fidelities (model, sequence length, batch, precision, prefill vs decode, tensor, parallelism)
• Bring up inference workloads from Hugging Face, PyTorch, papers, and vendor stacks (vLLM, SGLang, TensorRT-LLM, Triton)
• Measure competing GPUs and accelerators end-to-end, managing cloud or lab accounts, images, drivers, and run recipes
• Report TTFT, ITL, tokens/s, tokens/s per watt, and energy using nv-smI, DCGM, or equivalents
• Document discrepancies between RTL, model, and measured results, with configs/logs behind each
• Maintain a benchmark suite separating internal vs external results; route external claims for approval

Skills

Python programming

GPU performance analysis

Benchmark harness development

Linux proficiency

Education

BS or MS in Computer Engineering / Electrical Engineering / Computer Science

Tools

NVIDIA Nsight Systems

Nsight Compute

Hugging Face

TensorRT-LLM

Job description

Neurophos, Inc. is seeking a performance engineer to own benchmarking and energy metrics for the T100 optical inference accelerator.

You will define measurement methods, generate end-to-end results, and ensure reproducibility across models, fidelities, and hardware generations. The role requires 5+ years in GPU performance, with hands-on experience in roofline and analytical modeling, and the ability to work with cloud GPUs and ML stacks.