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
$108к$145кмедиана$196к
19 в выборке
На СНГ рынке: $5.3к/мес ($4к - $6к)
91
Крутая вакансия
развернуть
Чётко очерченная senior-роль в быстро развивающемся AI-стартапе даёт возможность работать с производительностью инференса и реальным оборудованием, хотя стартап-среда может быть менее предсказуемой.
?Темп стартапа
Актуальный стек
Передний край AI
Понятная зона ответственности
Мэтч & Сопровод
Для мэтча с этой вакансией нужен Plus
Подключить Plus
Описание вакансии
Текст:
Hirify AI/Оригинал
TL;DR
Senior Software Engineer, Inference Performance (AI): Building the software stack, APIs, kernels, libraries, and performance tooling that optimize multimodal generative AI models on custom silicon with an accent on hardware-aware inference, measurable performance, and scalable abstractions. Focus on profiling and eliminating bottlenecks across kernels, compilers, runtimes, and parallelism, building agentic performance loops, and carrying optimizations across future hardware generations.
Location: Munich, Germany; other location: München, Bayern, Germany
Company
hirify.global is a well-funded startup building custom silicon, hardware, and software for high-performance, low-power multimodal generative AI inference systems serving hyperscaler and neocloud data centers.
What you will do
Design APIs, kernel abstractions, and libraries in the Python SDK or Rust stack.
Optimize LLM and multimodal workloads on real silicon by profiling bottlenecks across kernels, compilers, runtimes, and parallelism.
Compare measured performance with analytical models and simulations, explain gaps, and implement fixes.
Build profilers, traces, counters, and benchmark harnesses for agentic performance optimization.
Define performance measurement, regression tracking, and reporting standards.
Design abstractions for future hardware generations and collaborate with hardware architects on workload impact.
Requirements
Experience optimizing AI workloads on GPUs or custom accelerators, with an understanding of compute utilization, memory hierarchy, latency hiding, synchronization, and multi-chip scaling.
Strong Python and/or systems programming skills in Rust, C++, or C, with experience owning substantial systems from design through production.
Strong engineering practices involving interfaces, abstractions, and testing.
End-to-end understanding of LLM inference, including prefill, decode, attention variants, KV caches, MoE, parallelism, and their compute, memory, and communication requirements.
Experience building with agents as part of the development workflow and understanding when human intervention is needed.
Experience designing APIs, libraries, or performance tooling adopted by other engineers.
Ability to form hypotheses, measure results, follow evidence across software and hardware boundaries, and own outcomes.
Nice to have
Embedded, bare-metal, or low-level runtime programming experience.
Experience with MLIR, compiler frameworks, or kernel DSLs such as Triton, Pallas, NKI, or CuTe.
Contributions to LLM serving stacks such as vLLM or SGLang, or experience delivering competitive end-to-end inference performance on a hardware target.
Culture & Benefits
Fast-paced startup environment focused on ambitious technical goals and engineering excellence.
Direct collaboration with hardware, compiler, runtime, and model engineers.
Comprehensive benefits for employees and families, flexible spending options, competitive compensation, and recognition programs.
Values include ownership, integrity, teamwork, diverse perspectives, and high-quality execution.
Equal opportunity employment is provided without discrimination based on protected characteristics.
Hiring process
Recruitment agencies should not contact hirify.global employees or leaders; unsolicited agency resumes are not accepted.
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 →