Principal GPU Inference Engineer & Team Lead (Remote)

Runware; United States; Remote; USD 300,000 - 500,000; Full time

Benefits offered by this job

Meaningful stock options

Remote-first with core hours

Generous PTO and holidays

Company retreats

Job summary

Runware, an AI inference platform, seeks a Principal Engineer to own model performance across its distributed fleet of GPUs. You will lead the Model Performance team, write critical kernels, and partner with infrastructure to optimize latency and cost while expanding the team around you.

This remote-first role offers strategic impact, ownership of the performance roadmap, and collaboration with NVIDIA and open‑source communities to land production-quality improvements across diverse models and

Qualifications

• 10+ years in GPU, HPC or DL systems engineering with production inference experience.
• Hands-on CUDA and GPU architecture expertise for low-latency deployment.
• Proven shipping performance from research to production on real hardware.
• Leads teams, hires and sets technical direction while remaining hands-on.

Responsibilities

• Own end-to-end inference performance for the fleet, covering kernels, serving, and model-level optimizations.
• Set the performance roadmap and decide models, hardware and techniques to tackle.
• Code and tune kernels, quantization, and parallelism on multi-GPU nodes.
• Build production-ready deployment pipelines for open models with low latency and cost.
• Shape serving stacks and decide what to adopt or fork, contributing upstream as needed.
• Own benchmarking and regression systems, with CI integration to prevent regressions.

Skills

CUDA kernel engineering

LLM inference optimization

multi-GPU systems

team leadership

roadmap ownership

cross-functional collaboration

Tools

TensorRT-LLM

vLLM

SGLang

Triton

Nsight (perf tools)

Job description

Runware, an AI inference platform, seeks a Principal Engineer to own model performance across its distributed fleet of GPUs. You will lead the Model Performance team, write critical kernels, and partner with infrastructure to optimize latency and cost while expanding the team around you.

This remote-first role offers strategic impact, ownership of the performance roadmap, and collaboration with NVIDIA and open‑source communities to land production-quality improvements across diverse models and