Вакансия из Hirify Global, списка международных tech-компаний Для мэтча и отклика нужен Plus
Пожаловаться
Зарплата и рынок
ЗП не указанаразвернуть
На похожих ролях по рынку
$170к$200кмедиана$227к
50 в выборке
На СНГ рынке: $3.7к/мес ($2.8к - $4.5к)
87
Крутая вакансия
развернуть
У роли чёткая ответственность за GPU-инфраструктуру в продакшене, а направление AI-вычислений сейчас особенно актуально; основные компромиссы - интенсивные релизные циклы и очень высокий порог опыта.
Интенсивные релизы
?Узкая экспертиза
Современный стек
Сфера AI-инфраструктуры
Понятная зона ответственности
Мэтч & Сопровод
Для мэтча с этой вакансией нужен Plus
Подключить Plus
Описание вакансии
Текст:
Hirify AI/Оригинал
TL;DR
Senior Staff Engineer - Infrastructure Systems (GPU) (NVIDIA/CUDA): Owning the GPU software stack across Runware's infrastructure, from firmware and kernel drivers to CUDA, containers, Kubernetes, and inference services, with an accent on fleet-wide qualification, diagnostics, and self-healing. Focus on bringing up new GPU generations, tracing failures across hardware and software layers, and ensuring reliable multi-tenant GPU serving at scale.
Location: Remote from California, United States. Remote-first work with core collaboration hours and in-person company retreats twice a year.
Company
hirify.global is building an AI supercomputer infrastructure spanning GPU nodes, Kubernetes, inference services, and a global network.
What you will do
Own versioned, reproducible GPU software baselines covering firmware, Linux kernel, NVIDIA drivers, CUDA, libraries, container tooling, and Kubernetes integration.
Qualify driver and CUDA releases across GPU generations, frameworks, and model families, identifying correctness, performance, and memory regressions before production rollout.
Root-cause GPU failures across hardware, firmware, drivers, runtimes, and applications, including Xid, ECC, PCIe, NVLink, thermal, memory, and bus failures.
Build node acceptance, burn-in, diagnostics, telemetry, fault classification, and automated drain, reset, and replacement workflows.
Own GPU exposure and isolation through container runtimes, device plugins, GPU Operator, MIG, sharing policies, bare metal, and Kubernetes.
Bring up new GPU generations, benchmark real inference workloads, work with NVIDIA and upstream projects, and use AI for triage, test generation, release qualification, and runbook automation.
Requirements
12+ years of experience building and operating systems software, including several years with production GPU software at scale.
Deep hands-on knowledge of the NVIDIA stack: kernel drivers and modules, CUDA runtime and libraries, NVML, DCGM, Container Toolkit, and version compatibility.
Experience taking GPU failures to root cause across hardware, firmware, drivers, runtimes, and applications, including Xid, ECC, row remapping, PCIe, NVLink, thermal, and memory faults.
Strong Linux internals knowledge covering kernels, device drivers, modules, PCIe, NUMA, IOMMU, cgroups, containers, perf, eBPF, and ftrace.
Daily coding experience with fluent Python and at least one of C, C++, Rust, or Go; sufficient CUDA proficiency to write tests and reproduce bugs.
Experience running GPUs under Kubernetes and on bare metal, including device plugins, GPU Operator, MIG, image builds, and large-scale provisioning.
Working knowledge of GPU inference frameworks such as PyTorch, TensorRT, and ONNX Runtime, including memory allocation, compilation, caching, and multi-GPU execution.
Clear technical writing for design documents, postmortems, and actionable ticket updates.
Daily use of AI tools to accelerate output, with a clear view of how AI-enabled small teams should operate.
Ability to own outcomes, prioritize in ambiguity, and remain accountable in a fast-moving environment.
Nice to have
GPU driver, CUDA library, or GPU systems experience at a GPU vendor, cloud provider, or HPC centre.
Blackwell hardware, including B200, GB200/GB300 NVL72, NVLink, NVSwitch, and liquid-cooled racks.
Multi-GPU and multi-node execution with NCCL, GPUDirect, RDMA over RoCE, or InfiniBand.
Firmware qualification across BIOS, BMC, NIC, and GPU, plus OEM and ODM acceptance testing.
Inference serving internals involving TensorRT-LLM, vLLM, SGLang, or custom diffusion and video pipelines.
Non-NVIDIA accelerator software stacks, production open-source projects, multi-tenant isolation, confidential computing, or 24x7 services with published service levels.
Culture & Benefits
Remote-first work with flexible hours outside core collaboration blocks.
In-person company retreats twice a year.
Fast, intense release cycles followed by planned downtime.
Paid vacation, sick days, public holidays, maternity, paternity, and caregiver leave.
Meaningful stock options and schedule autonomy.
Будьте осторожны
Если работодатель просит войти в их систему, используя iCloud/Google, прислать код/пароль, запустить код/ПО, не делайте этого. Это мошенники. Обязательно жмите «Пожаловаться» или пишите в поддержку. Подробнее в гайде →