Вакансия из Hirify Global, списка международных tech-компаний Для мэтча и отклика нужен Plus
Пожаловаться
Зарплата и рынок
развернуть
$180к-450к/год
$170к$210кмедиана$270к
По похожим ролям (47) ~ $210к/год
На СНГ рынке: $4к/мес ($3.6к - $5.9к)
91
Крутая вакансия
развернуть
Это хорошо очерченная роль в передовом мультимодальном ИИ с практической работой над обучением моделей и GPU-кластерами; главный вопрос - насколько молодой компании удастся превратить амбициозную идею продукта в устойчивый бизнес.
Передовой ИИ
Понятные задачи
?Риск молодого бизнеса
Мэтч & Сопровод
Для мэтча с этой вакансией нужен Plus
Подключить Plus
Описание вакансии
Текст:
Hirify AI/Оригинал
TL;DR
Member of Technical Staff, Mid-training (AI) (multimodal foundation models): Developing mid-training recipes, synthetic data pipelines, and long-context data strategies for models spanning text, audio, code, and other modalities with an accent on data mixtures, learning-rate schedules, distillation, and checkpoint evaluation. Focus on designing controlled experiments, debugging distributed GPU training, and measuring how training changes transfer to later model adaptation.
Location: San Jose, United States
Salary: $180,000–$450,000 base salary annually
Company
hirify.global is an artificial intelligence company developing personalized, proactive, multimodal intelligence and next-generation hardware for interaction through speech, text, vision, and persistent memory.
What you will do
Develop mid-training recipes using high-quality and synthetic data, learning-rate decay schedules, and changing data mixtures.
Build synthetic data pipelines with code execution tests, math-answer checks, reasoning-trace filters, generation, verification, and rejection sampling.
Develop long-context data recipes covering document packing, sequence construction, and context-extension evaluation.
Distill capabilities from larger models into smaller models and evaluate their effects on model quality and data diversity.
Evaluate mid-training checkpoints with post-training researchers, measure their predictive value for later model quality, and select checkpoints for further training.
Run controlled experiments across learning-rate schedules, mixture weights, data ordering, and text, code, and multimodal training data.
Requirements
Experience with continued pretraining or late-stage training, including learning-rate decay, high-quality data mixtures, or context extension.
Experience building synthetic data pipelines and evaluating effects on model quality and data diversity.
Strong experimental judgment, including controlled ablations and assessing whether improvements transfer to larger runs.
Experience implementing and debugging distributed training workloads across GPU clusters.
Experience designing evaluations and identifying benchmark contamination, prompt sensitivity, and formatting effects.
Strong engineering skills in implementing training methods and building reliable data and experimentation pipelines.
A record of improving model capabilities or training methods through research, datasets, systems, or production work.
Nice to have
Experience with distillation, rejection sampling, or model-based data generation for reasoning, code, or math.
Experience with long-context training, including RoPE scaling, context parallelism, or long-document datasets.
Experience with post-training and evaluating pretrained checkpoints during subsequent adaptation.
Experience training or distilling smaller models for constrained inference budgets.
Experience with audio, speech, image, video, or interleaved multimodal datasets.
Experience detecting benchmark contamination and removing overlapping training examples.
Culture & Benefits
Full-time position with a US annual base salary range of $180,000–$450,000.
Total compensation may include additional components and benefits depending on the role.
Compensation varies according to job-related knowledge, skills, and experience.
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 →