Description

We are the Intelligent Search team, building a search service that answers user queries in natural language.

Team Mission: Break the barrier between the static knowledge of a language model and the constantly changing world. We provide GigaChat with access to up-to-date information so users get accurate answers to any questions, including questions about the latest news and events.

What we are working on:

• Development of the search core: We create and optimize the engine that understands what exactly the user needs and finds it in the sea of data

• Integration with GigaChat: We ensure the seamless operation of our search within Sber's main LLM

• Multi-platform: We make our functionality available to users on the web platform, in messengers, and on smart devices

• Solving complex tasks: We overcome challenges related to processing, structuring, and delivering data in a format that is "digestible" for a language model

Our technology is a key component that:

• Allows the LLM to go beyond the boundaries of its training dataset

• Provides answers based on verified and fresh data

In our project, we use the most modern approaches to solving the Information Retrieval task, conduct experiments to improve search result ranking using machine learning methods, launch and analyze A/B test results.

If you have always wanted to solve tasks at the intersection of science and technology, work in the field of information retrieval, engage not only in research tasks from the NLP field but also independently implement and deploy modern algorithms from recent scientific articles, then you are our ideal candidate!

Project technology stack: Python, OpenSearch/ElasticSearch, Transformers, PyTorch, Airflow, PostgreSQL, CatBoost, Hadoop, PySpark.

Responsibilities

• participate in the development of the search engine
• develop and optimize machine learning algorithms and models
• train Transformer-based models
• conduct research and experiments to improve search quality
• deploy ML solutions to production
• conduct A/B tests
• experience in the field of NLP / ranking / search / recommendations of at least three years
• good knowledge of classic data structures and algorithms
• fundamental understanding of how standard machine learning algorithms work
• experience working with modern language models - transformers
• proficient working knowledge of Python
• ability to write readable and maintainable code.

Requirements

• experience in the field of NLP / ranking / search / recommendations of at least three years
• good knowledge of classic data structures and algorithms
• fundamental understanding of how standard machine learning algorithms work
• experience working with modern language models - transformers
• proficient working knowledge of Python
• ability to write readable and maintainable code.

Will be a plus:

• experience working with ElasticSearch, OpenSearch or another search engine as a search tool, not just a log storage
• experience working with Retrieval Augmented Generation
• knowledge of approximate nearest neighbor search algorithms
• knowledge of the basics of C++ and Java development
• experience in implementing and deploying machine learning models
• ability to work with crowdsourcing platforms
• familiarity with PySpark.

Conditions

• comfortable modern office near Kutuzovskaya metro station
• hybrid work format
• annual salary review, annual bonus
• corporate gym and recreation areas
• training system for professional and career development
• extended voluntary health insurance from the first day of work and insurance for family
• employee mortgage program
• free SberPrime+ subscription, discounts on partner company products
• referral bonus for recommending friends to the Sber team.