Чётко очерченная роль в ведущей AI-компании с прямой работой над обучением и безопасностью передовых моделей в быстроразвивающейся области.
На переднем крае AI
Понятная роль
Работа с передовыми моделями
Выше рынка
Мэтч & Сопровод
Для мэтча с этой вакансией нужен Plus
Подключить Plus
Описание вакансии
Текст:
Hirify AI/Оригинал
TL;DR
Researcher, Cyber Safety Training (AI): Training and evaluating frontier AI models to provide legitimate cybersecurity assistance while recognizing potentially harmful requests, with an accent on post-training, reinforcement learning, and realistic cyber safety evaluations. Focus on designing complex evaluations, integrating robust safety methods into large model runs, and verifying behavioral improvements at scale.
Location: San Francisco, United States
Salary: $380K–$500K plus equity
Company
OpenAI is an AI research and deployment company developing general-purpose AI systems intended to benefit humanity and be deployed safely.
What you will do
Work with cybersecurity researchers to identify how context changes the assistance a model should provide.
Design complex and realistic cyber safety evaluations.
Research post-training techniques that improve the cyber safety of frontier models.
Collaborate with the core post-training team to integrate promising methods into large model runs and verify safety improvements at scale.
Coordinate with trust and safety, legal, policy, and research teams to meet high safety standards.
Requirements
Hands-on experience training or fine-tuning machine learning models, including describing an experiment whose results changed the next training approach.
Practical security judgment gained through industry work, a research project, collaboration with a security team, or capture-the-flag competitions.
Ability to build and debug training-experiment code, from data preparation through verification of changes in model behavior.
Ability to work through ambiguous cases with domain experts and translate findings into concrete model-response decisions.
Nice to have
Reinforcement learning experience combined with motivation to develop substantial cybersecurity expertise.
Cybersecurity experience combined with an interest in deepening post-training expertise.
Culture & Benefits
Full-time employment in OpenAI's Safety Research and Safety Training organization.
Work spans longer-term safety research and production post-training for models across types, modalities, and product surfaces.
Cross-functional collaboration with research, trust and safety, legal, and policy teams.
Equity is included in the compensation offer.
OpenAI provides reasonable accommodations for applicants with disabilities and considers qualified applicants with arrest or conviction records in accordance with applicable law.
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 →