Search is the core of Yandex. Every user query is a task of selecting the perfect search results from billions of documents. Everything depends on the quality of this solution: trust in Search, the success of Alice, and the future of generative answers on the results page.

We are developing ranking—one of the largest ML systems in the country. We are looking for a leader who will take responsibility for the quality of search for 100+ million users. In this role, you will define what Search will become: from metrics and signals to model architecture and solutions that reach production.

Why this role is important:

• You directly influence Yandex's key product.
• You work with big data, complex ML models, and real user impact.
• You make architectural and product decisions that scale across the entire country.

If it's important to you that ML doesn't work in a vacuum but changes the product and user experience—join us.

What tasks await you

Developing the ranking core You will improve the main ranking model, work with training datasets: combat biases, clickjacking, and noise. You will need to design and select targets aligned with the user quality of results, and combine signals and targets to ensure consistently good results across all key metrics.

Leadership and management You will lead a team of ML developers: set goals, prioritize directions, decompose major initiatives, and bring them to production. You will be responsible for the quality of the team's solutions—from idea and experiment to stable operation in production—and develop the team: help them grow technically and in product sense.

Cross-team collaboration Search is a complex system where ML does not exist separately from other areas. You will work closely with analysts and backend teams, jointly design solutions, choose optimal approaches, and control implementation quality, participate in discussions of metrics, experiments, and product trade-offs.

Models for specific quality aspects When general signals are insufficient, you will analyze ranking weaknesses, design improvement architecture, and together with the team, train and implement specialized ML/DL models, such as document quality assessment, grouping of documents from the same site, detection of duplicates and near-duplicates.

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

We expect you to

• Have managed a team of ML developers and been responsible for results
• Have a deep understanding of classical ML and have applied it in production
• Are not afraid to work hands-on: prototyping, testing hypotheses, analyzing data
• Know how to critically evaluate metrics and model results
• Want to build a very high-quality product, not just improve numbers
• Are interested in tasks at the intersection of ML, analytics, and product