Чётко очерченная роль в ведущей AI-компании сочетает причинно-следственный анализ с прямым влиянием на безопасность и защиту продуктов.
Выше рынка
Безопасность ИИ
Современный стек
Понятная роль
Мэтч & Сопровод
Для мэтча с этой вакансией нужен Plus
Подключить Plus
Описание вакансии
Текст:
Hirify AI/Оригинал
TL;DR
Data Scientist, Integrity (AI): Investigating fraud, abuse, and harmful misuse across ChatGPT, Codex, and API products with an accent on anomaly detection, experimentation, causal inference, and measuring customer impact. Focus on evaluating detection systems, designing controls and guardrails, and building analyses that balance prevented harm against legitimate-user friction, revenue, compute costs, and review workload.
Location: San Francisco, United States; hybrid
Salary: $265K–$380K annually plus equity
Company
OpenAI is an AI research and deployment company developing general-purpose AI systems and products including ChatGPT, Codex, and the API.
What you will do
Investigate anomalous activity across products, estimate abuse prevalence and impact, and recommend intervention priorities.
Define outcome metrics, guardrails, baselines, and monitoring for abuse, customer friction, and operational performance.
Build and evaluate detection rules, classifiers, and human review workflows using labeled samples and review outcomes.
Design randomized holdouts and apply quasi-experimental or observational methods to measure fraud losses prevented and effects on legitimate users, revenue, compute costs, and review workload.
Trace failures in signals, review decisions, and enforcement actions, then build reusable datasets, analyses, and investigation tools with Product, Engineering, Research, Policy, and Operations.
Requirements
Use AI tools to investigate large, messy datasets, test hypotheses, and automate recurring analyses.
Fluency in SQL and Python, with the ability to verify AI-generated queries, code, and conclusions against underlying data.
Strong statistical judgment and experience with experimentation, causal inference, or observational analysis.
Ability to reason from incomplete or biased labels and explain how uncertainty affects decisions.
Ownership of open-ended analytical problems from defining measurements through recommending actions and checking outcomes.
Clear communication with technical and nontechnical partners, including explanation of recommendation tradeoffs.
Nice to have
Experience in integrity, trust and safety, fraud or risk analytics, security analytics, or applied ML evaluation.
Experience evaluating detection systems, human review workflows, or LLM-based classifiers.
Culture & Benefits
Hybrid work arrangement in San Francisco.
Equity is included in the compensation package.
OpenAI is committed to equal opportunity, reasonable accommodations, and consideration of 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 →