Get a reply from this employer
Your resume and cover letter, tailored to exactly what they’re hiring for.
Runware in California seeks a hands-on, senior contributor who owns the full GPU software stack—from firmware and kernel up through drivers, CUDA libraries, containers and Kubernetes. The ideal candidate has a decade of systems work and can trace faults to their root cause while writing code daily.
You will diagnose, triage, and fix complex GPU issues, document findings clearly, and help scale the platform for model inference.
This is a hands-on, top-tier individual contributor role that owns the full GPU software stack on every node, from firmware and kernel up through drivers, CUDA libraries, containers and Kubernetes. A strong candidate has more than a decade of systems work, years of it on production GPUs, and can trace a fault such as an Xid error or a GPU dropping off the bus to its true cause. They also write code daily, explain findings clearly in writing, and pick the most important problem without being told.
AI tools can speed up the repetitive parts, such as sorting through logs and Xid reports, generating tests, qualifying new releases and turning fixes into runbooks. The posting expects the person to use these tools daily and to rethink how a small team works. What stays human is judgment on ambiguous failures, proving whether the cause is hardware, firmware, driver or application, and working with NVIDIA and other teams to get problems fixed.
The role sits at the base of an inference platform serving many model types, so it offers close exposure to how image, video, audio and language models actually use GPUs. Someone could gain experience benchmarking new GPU generations with real workloads and building AI agents for fault triage and release checks. To prepare, practise debugging GPU faults on a home or lab machine and try an agent that reads driver logs and suggests causes.
This is our own read, written from the employer's posting and the company's Spotlight. The employer's posting is the source of truth for the role.
High-performance inference platform providing serverless and dedicated GPU infrastructure across thousands of machine learning models.