IT B2C is the largest ecosystem in Sber. We are over 8000 people in 18 cities across Russia. We are developing and improving retail solutions, helping to make the Bank's services more accessible, secure, and convenient.
We are waiting for you!
About Us:
We develop agent RAG systems that help the company improve customer experience through personalized solutions.
We already have a working product that is in production and brings value. Currently, the team is actively working on new applications of the technology and improving the quality of existing ones.
Responsibilities
Under the guidance of a mentor, conduct experiments to improve agent RAG systems:
- improve quality measurement methods: annotation, benchmark collection, and llm-as-a-judge.
- improve search by training models and the quality of sources themselves.
- optimize and improve response generation.
- together with engineers from the team, deploy new models/approaches to production.
- develop integrations for reaching new customer touchpoints.
- conduct code reviews, monitor code quality, write tests.
- develop and implement new services with ML models.
Requirements
- knowledge of the fundamentals of classical ML and NLP/LLM: classic tasks and architectures/approaches for their solutions, metrics, model training process.
- confident knowledge of Python and various ML/DL frameworks: PyTorch, Transformers, pymorphy, etc.
- knowledge of RAG fundamentals: understanding what tasks it is used for, what problems it solves, and how to implement a basic approach without using langchain.
- proficiency in a basic set of development skills: bash, git, etc.
- pet projects/competitions for solving NLP tasks: text classification, summarization, search, etc.
Conditions
- office-based work: Moscow, Kutuzovsky prospect, 32, bldg. 1
- annual salary review and annual bonus.
- over 400 educational programs at SberUniversity for your development.
- gym and relaxation areas.
- extended voluntary medical insurance, preferential insurance for family members.
- flexible mortgage discount equal to 1/3 of the Central Bank's key rate.
- Prime subscription with the possibility of shared use by three close individuals.
- reward for recommending friends to the Sber team.
