Description

About us:

We are the B2C Digital Assistant team. We create a product that helps tens of millions of our clients daily to use the Bank's and Ecosystem's services and products at a new level.

Currently, our team is developing new functionality that allows clients to solve not only everyday but also more complex, comprehensive tasks and goals, which may consist of many interconnected steps and be distributed over time.

To solve the task, we have at our disposal a digital platform and a wide range of tools, represented by a multi-agent system based on the LLM GigaChat, integrated with numerous bank services.

We offer:

· Opportunity to create and develop an innovative product using cutting-edge technologies and setting trends in the industry.

· Work in a cross-functional team of strong specialists.

· Wide scope for the development and implementation of your ideas and expertise.

· Modern technology stack and advanced IT infrastructure.

Responsibilities

• Designing and developing new platform tools and agents and expanding the functionality of existing ones.
• Conducting deep analysis of data and user scenarios of the multi-agent system, developing functional requirements for new specialized agents.
• Developing and improving platform mechanisms such as routing, orchestration, planning, context and memory management, insight extraction systems.
• Developing a comprehensive quality metrics system for both individual system components and entire system components.
• Creating automation pipelines for monitoring and visualizing metrics, including using the LLM-as-a-Judge approach.

Requirements

· Higher technical or mathematical education.

· Experience in machine learning and data analysis from three years.

· Deep understanding of the principles and architectures of LLMs, NLP, and classical ML algorithms, knowledge of mathematical statistics and probability theory.

· Proficiency in key tools and frameworks for ML development, analytics, and big data processing, working with LLM agents (pandas, numpy, sklearn, pytorch, langchain, langgraph, pyspark).

· Confident practical skills in working with LLMs and prompt engineering.

· Understanding the principles of building industrial microservice architectures, multi-agent systems, and LLM applications.

Conditions

· Hybrid work mode, comfortable modern office near Kutuzovskaya metro station.

· Annual salary review, annual bonus.

· Corporate gym and relaxation areas.

· More than 400 educational programs from SberUniversity for professional and career development.

· Extended private health insurance (DMS), preferential insurance for family, and corporate pension program.

· Flexible discount on mortgage loan, equal to 1/3 of the Central Bank's key rate.

· Free SberPrime+ subscription, discounts on products of partner companies.

· Reward for recommending friends to the Sber team.