ML Systems Engineer, Model Architectures (Mid/Senior/Principal) (AI)
Контакты
160 000 - 350 000$
Формат работы
remote (только USA)
Тип работы
fulltime
Грейд
middle/senior
Английский
b2
Страна
US
aipythonc++transformersmachine learningpytorchruststartupquantizationinferencemixture of experts
Вакансия из Hirify Global, списка международных tech-компаний Для мэтча и отклика нужен Plus
Пожаловаться
Зарплата и рынок
развернуть
$160к-350к/год
$210к$250кмедиана$320к
По похожим ролям (49) ~ $250к/год
На СНГ рынке: $5.3к/мес ($3.8к - $6к)
91
Крутая вакансия
развернуть
Чёткая практическая роль на стыке передовых AI-моделей и специализированных ускорителей, причём результаты напрямую влияют на дизайн чипов; главный компромисс - работа в стартапе.
Риски стартапа
Актуальный стек
Понятные задачи
AI-железо
Хороший пакет
Только США
Мэтч & Сопровод
Для мэтча с этой вакансией нужен Plus
Подключить Plus
Описание вакансии
Текст:
Hirify AI/Оригинал
TL;DR
ML Systems Engineer, Model Architectures (Mid/Senior/Principal) (AI): Implementing and validating new transformer model architectures in simulators for low-latency AI inference accelerators with an accent on attention variants, mixture-of-experts, embeddings, quantization, and bit-exact numerical correctness. Focus on evaluating compute, memory, and data-movement requirements before hardware exists, bringing customer models up in simulation, and automating model bring-up across architecture, compiler, SDK, and customer teams.
Location: Remote (United States)
Base salary: Mid: $160,000–$220,000; Senior: $220,000–$270,000; Principal: $270,000–$350,000 per year.
Company
hirify.global develops AI inference accelerators for low-latency, high-throughput large language model inference, including Atlas systems in production and the Asimov ASIC and Titan system.
What you will do
Implement operators, numerics, and memory layouts for new model architectures in a simulator with bit-exact reference validation.
Extend inference coverage across prefill and decode, embeddings, grouped-query, gated, linear and sparse attention, mixture-of-experts routing, and quantized formats.
Maintain reference comparisons and regression coverage as simulator and hardware implementations evolve.
Evaluate new architectures before hardware exists, identifying compute, memory, data-movement, and chip-design requirements.
Bring customer models up in simulation ahead of emulator and hardware runs, owning turnaround time.
Partner with agentic tooling, chip architecture, compiler, SDK, and customer teams to automate model bring-up from specification through simulator coverage.
Requirements
4+ years of experience as an ML systems or inference engineer.
Strong operator-level understanding of transformer internals, including attention variants, mixture-of-experts, embeddings, KV cache, and INT8 and FP8 quantization.
Demonstrated experience implementing or porting model inference by hand in PyTorch, llama.cpp, vLLM, TensorRT-LLM, or a custom runtime.
Hands-on experience debugging numerical mismatches against a reference down to the operator and bit level.
Ability to read model architecture papers and turn them into working, verified implementations.
Strong Rust skills, or expert-level C/C++ skills with willingness to work in Rust daily, plus strong Python skills for model and reference work.
Excellent communication skills for explaining hardware requirements of new architectures and current design limitations.
Must currently be authorized to work in the United States.
Nice to have
Experience mapping models onto accelerator hardware or writing kernels for a non-GPU architecture.
Experience serving as the technical point of contact for customer model requests.
Culture & Benefits
Remote-first work with a company-provided computer and home office setup.
Company-paid medical, dental, vision, life, and disability coverage, including dependent medical coverage.
Unlimited paid time off and 13 paid company holidays.
Competitive salary, equity, and 401(k) matching eligible from day one.
No new visa sponsorship; H-1B transfers may be facilitated for eligible candidates.
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 →