ML Developer for Retail Risks at Yandex Bank

Our team develops and implements ML solutions for managing retail risks in Yandex Fintech. Decisions based on our models are made both in Russia and internationally.

The team is responsible for the entire model lifecycle: from domain analysis and client discussions to creating ML models, deploying them into production, and further support. This approach allows us to deeply understand tasks and influence product development, focusing on business metrics.

What tasks await you

Building scoring models for Yandex fintech products Scoring models allow for assessing the creditworthiness and financial risks of users of Yandex's fintech services. This enables making informed decisions about providing personalized financial services, minimizing losses from non-payments, and increasing product accessibility for reliable customers. Development is conducted for both credit products on the Russian market and for the international direction.

Building models for predicting user income, actions, and behavior By analyzing customer activity in Yandex services, we predict their income and financial behavior. This allows us not to guess needs, but to create offers perfectly suited to each user, considering their financial profile and life situation.

Designing models, experiments, and A/B tests Before developing a model, we thoroughly work out its design and usage scenarios. This helps ensure correct and effective application in business processes, provides necessary flexibility, and the ability to scale solutions. We design experiments and A/B tests, which allow us to collect the necessary data for training models and measure the impact of models on real business metrics.

Analyzing the effectiveness of new data sources and applying them in models When creating models, there is an opportunity to use large arrays of various data from both internal Yandex services and partners. By analyzing new information sources, one can significantly increase both model quality and the level of service of products.

Deploying models to production, maintaining their stability, and improving them After development and testing, models are deployed into the production environment, where key performance indicators can be tracked. You will have the opportunity to continue developing already deployed models, testing new hypotheses and modern approaches.

More about ML at Yandex — in the channel Yandex for ML

We expect you to

• Know how to analyze data in Python
• Know how to write SQL queries
• Understand the principles of classical ML
• Have developed models and presented them to the client or implemented them in production
• Understand the basics of probability theory, mathematical statistics, and A/B tests
• Be willing to engage in all stages of model creation: from discussing requirements to supporting finished solutions