Senior Researcher

About the company

The Research Center is engaged in creating SoTA artificial intelligence technologies for analyzing financial data and their application for the AI transformation of the Risk Block. Our task is to develop and train a multimodal reasoning model that combines textual data and digital customer footprints to improve the quality of decisions for banking tasks.

Your tasks

• perform generation and cleaning of synthetic data using LLM and LLM-as-a-Judge approaches
• train a large multimodal LLM (text + digital customer footprints)
• apply RL approaches to improve the reasoning capabilities of the model
• perform validation of the developed solution on modern benchmarks in terms of accuracy and reasoning correctness
• evaluate solutions on real-world data.

Requirements

Requirements

• 3+ years of experience
• deep knowledge of ML/DL (transformers, large language models)
• confident experience in training LLM models
• proficient knowledge of Python: torch, transformers, peft
• English B2+.

Will be a plus

• experience working with multimodal models
• knowledge of RL methods for LLM (RLHF, DPO, GRPO, RLVR, GSPO)
• experience in distributed training of large models (FSDP, ZeRO, Mixed-precision, Tensor/Pipeline Parallelism)
• publications in top journals (Q1) or conferences (A-A*)
• general understanding of modern AI tools and their capabilities.

Requirements Higher education; Experience: 5 years

What we offer

• comfortable modern office near Kutuzovsky pr. or Volgogradsky prospekt metro stations
• work format - office
• annual salary review and annual bonus
• corporate gym and relaxation areas
• more than 400 educational programs from SberUniversity for professional and career development
• adaptation program and manager assistance at the start (for Junior level vacancies)
• extended VHI, discounted insurance for family, and a corporate pension program
• flexible mortgage discount equal to 1/3 of the Central Bank key rate
• Prime subscription with the possibility of shared use among three close ones
• reward for recommending friends to the Sber team.

Other

Hardware and data

• dedicated cluster of 500+ GPU A100 80GB
• Spark/Hadoop for working with Big Data
• access to all banking data necessary to build the solution.