Model-Performance Principal Engineer, Remote & Stock Options

Runware; San Francisco (CA); Remote; USD 300,000 - 430,000; Full time

Benefits offered by this job

Stock options

Remote-first culture

Generous PTO

Parental leave

Company retreats

Job summary

Runware is seeking a Principal Engineer, Model Performance to own end-to-end inference speed and cost across thousands of GPUs and tens of thousands of models. You will define the performance roadmap, implement high‑efficiency kernels, and push production‑grade optimizations across the stack.

You will collaborate with the Sonic Inference Engine, infrastructure and hardware teams, and industry partners to keep latency and cost at the Pareto frontier while growing a high‑impact team of engineers.

Qualifications

• 10+ years in GPU, HPC or deep learning systems engineering, optimizing inference performance at scale.

Responsibilities

• Own inference performance end to end for the fleet across image, video, audio, 3D and LLM workloads on multiple GPU generations.
• Write and tune CUDA kernels, fuse attention/MLP paths, and implement quantization formats with quality.
• Lead the Model Performance team, hire, mentor and set technical direction.

Skills

GPU/HPC engineering

CUDA programming

Python

C++

TensorRT-LLM

Benchmarking

Team leadership

Tools

TensorRT-LLM

vLLM

NCCL

Triton

Job description

Runware is seeking a Principal Engineer, Model Performance to own end-to-end inference speed and cost across thousands of GPUs and tens of thousands of models. You will define the performance roadmap, implement high‑efficiency kernels, and push production‑grade optimizations across the stack.

You will collaborate with the Sonic Inference Engine, infrastructure and hardware teams, and industry partners to keep latency and cost at the Pareto frontier while growing a high‑impact team of engineers.