anti-fraud engineer

Toloka; United States; Remote; USD 90,000 - 120,000; Full time

Benefits offered by this job

Remote or hybrid employment

IT setup and home office allowances

Competitive compensation with base, 1%

Job summary

Toloka — глобальная платформа для сбора и анализа данных, помогающая развивать фронтовые модели ИИ. Команда занимается борьбой с мошенничеством и проектирует защитные механизмы, включая правила, модели и инфраструктуру обработки данных.

Роль предполагает работу с большими данными, проектирование метрик и построение систем мониторинга, а также сотрудничество на международном уровне в гибридном формате работы.

Qualifications

• Вы формулируете неясную задачу, подбираете необходимые метрики и делаете обоснованные выводы из неструктурированных данных.
• Вы генерируете идеи на основе данных и умеете их корректно тестировать.
• Вы пишете уверенные, эффективные запросы для больших датасетов.

Responsibilities

• Изучать данные, находить новые паттерны мошенничества, а также инсайты и способы предотвращения.
• Разрабатывать метрики защиты и строить панели для их отображения.
• Преобразовывать гипотезы в решения: правила, алгоритмы и модели.

Skills

Data analysis

SQL

Python

Airflow

Machine learning

LLMs

Anomaly detection

Fraud detection

Tools

Airflow

ClickHouse

Databricks

Job description

Описание:

At Toloka AI we create data that powers leading GenAI models and innovations. We work with frontier labs, big tech, renowned AI startups, enterprises and non-profit research organizations worldwide. Our team builds the systems that detect and stop fraud on the Toloka platform. Fraud here is a moving target: bad actors adapt quickly, test our defenses, and increasingly use automation and LLMs to get around them.

Задачи:

• Explore data to find new fraud patterns, anomalies, and insights, and to understand how fraudsters operate.
• Design metrics that show how our protection works, and build dashboards and charts that make them visible.
• Generate hypotheses from data and test them rigorously.
• Turn validated hypotheses into production solutions: rules and heuristics, ML models, and LLM-based approaches.
• Build and maintain data pipelines (SQL, Python, Airflow) that keep our defenses running and up to date.
• Think like an attacker.
• Anticipate how a new defense could be bypassed, and design it to hold up from day one.
• Own your area end to end: from the first insight to a deployed solution and its ongoing monitoring.
• Help shape anti-fraud rules and policies, including how we define and classify fraud.

Требования:

• You can frame a vague problem, choose the right metrics, and draw sound conclusions from messy data.
• You generate ideas from data and know how to test them properly.
• You write confident, efficient queries on large datasets.
• You can build features, train and evaluate models, and ship them to production.
• You have working experience with Airflow or similar orchestration tools.
• You understand the trade-offs of protection: false positives, impact on honest users, and cost.

Будет плюсом:

• ClickHouse experience is a plus
• Practical experience with LLMs is a plus
• Hands-on experience in anti-fraud, trust & safety, risk, or abuse prevention (a big plus)
• Experience with anomaly detection, graph analysis, or behavioral modeling
• Experience applying LLMs to analytical or classification tasks
• Experience with Databricks
• Experience with crowdsourcing or data annotation platforms

Условия:

• You will be part of an international, dynamic environment that drives innovation and sets new standards in the AI and technology sector.
• Competitive compensation package including base salary, bonus, and ESOP.
• Flexibility: we offer remote or hybrid employment.
• Paid PTO and benefits will vary depending on location.
• IT setup and home office allowances.
• Personal and career development: we support employee's ambition for professional development and encourage them to implement their ideas in their projects;