Влиятельная роль в быстрорастущей AI-инфраструктурной компании с передовым стеком, хотя базовая зарплата чуть ниже медианы рынка для такого уровня специализации.
Зарплата ниже рынка
Ограничение по гражданству
Топовый стек
Хайповый домен
Быстрый рост
Понятный скоуп
Мэтч & Сопровод
Для мэтча с этой вакансией нужен Plus
Подключить Plus
Описание вакансии
Текст:
Hirify AI/Оригинал
TL;DR
Senior Software Engineer (GPU Kernel Optimization): Authoring and optimizing CUDA kernels for LLM inference with an accent on throughput, latency, and hardware utilization. Focus on exploiting tensor cores, tuning memory hierarchy, and implementing industry-leading benchmarking like MLPerf.
Location: Sunnyvale, CA or Bellevue, WA. Must be a U.S. person (citizen, green card holder, etc.) for export control compliance.
Salary: $182,000 – $242,000
Company
CoreWeave is an AI-cloud provider delivering high-performance GPU infrastructure for AI/ML, visual effects, and real-time inference.
What you will do
Author, profile, and optimize CUDA kernels (GEMMs, attention, MoE) on the critical path of LLM inference.
Optimize hardware usage, including tensor cores, memory coalescing, and shared-memory/register usage.
Prototype and ship kernels using DSLs and compilers to ensure maximum performance.
Build reproducible microbenchmarks and roofline analyses to validate end-to-end latency and throughput gains.
Implement and maintain benchmarking workflows for MLPerf Inference and Training.
Lead design reviews, mentor junior engineers, and drive overall architectural standards.
Requirements
5+ years of experience in high-performance computing, GPU software, or performance-critical systems.
Hands-on CUDA experience is required, including fluency with programming and memory models.
Deep understanding of GPU architecture (tensor cores, warp occupancy, NVLink/PCIe) and profiling with Nsight.
Strong proficiency in C++ and Python for low-level, performance-sensitive code.
Familiarity with model-serving stacks such as vLLM, TensorRT-LLM, or SGLang.
Must be a U.S. person (citizen, lawful permanent resident, refugee, or asylee) for export control compliance.
Nice to have
Experience with Triton, Mojo, or CuTe DSL for kernel authoring.
Knowledge of JAX/Pallas or AMD GPU experience (HIP/ROCm).
Expertise in NCCL and collective-communication performance.
Contributions to OSS projects like vLLM, SGLang, PyTorch, or Triton.
Experience with Kubernetes at production scale or Slurm for scheduling large GPU jobs.
Culture & Benefits
100% company-paid medical, dental, and vision insurance.
401(k) with generous employer match and Employee Stock Purchase Program (ESPP).
Flexible PTO and company-paid life and disability insurance.
Catered daily lunch in office and data center locations.
Comprehensive family support, including paid parental leave and childcare support via Kinside.
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 →