Description
We develop AI assistants for automating customer support within Sber's Ecosystem (OKKO, Samokat, EApteka). Our ML and LLM-based solutions independently handle inquiries, reducing the operator load and serving up to 6 million requests per month. As part of the team, you will be responsible for the full cycle of creating AI/ML solutions — from idea generation to deployment into a high-load production environment. We are looking for an NLP Data Scientist to develop and enhance solutions with a focus on working with LLM models, primarily GigaChat.
Responsibilities
- Development and implementation of LLM models to solve NLU business tasks (SFT, RAG (Retrieval Augmented Generation), Agents, Summarization, Text Ranking, Text Matching, Language Modeling)
- Fine-tuning of LLM models
- Development of approaches and processes for data annotation to evaluate the quality of LLM performance
- Forming and working with ML pipelines: data handling, training/fine-tuning NLP models, solution quality assessment, support/automation of solutions
- Optimization of model performance for the industrial environment on CPU/GPU
- Collaboration with the team of business representatives and DS developers
- Organizing validation and generating hypotheses for solving technical and business tasks.
Requirements
- Experience in development with python, numpy, sklearn, pandas + text data processing libraries
- Experience with Pytorch for building DL text models
- Experience with LangChain/LangGraph libraries
- Practical experience working with LLMs via API
- Excellent theoretical knowledge of classical and neural network NLP, including LLMs
- Experience in fine-tuning classic transformers and LLMs
- Practical experience, experiments, and deployment of LLM solutions into PROD
- Experience in prompt-engineering
- Knowledge of sql
- Linux, Git.
Will be a plus:
- Working with Hadoop tools (HDFS, Hive), Spark
- Experience in setting up and conducting a/b tests
- Experience with distributed training, deep knowledge of GPU architecture.
Conditions
- Comfortable modern office near Kutuzovskaya metro station
- Work format - hybrid is possible after the probationary period
- Annual salary review, annual bonus
- Corporate gym and relaxation areas
- Access to over 400 educational programs from SberUniversity for professional and career development
- Comprehensive VHI, preferential family insurance, and corporate pension program
- Flexible mortgage discount equal to 1/3 of the Central Bank's key rate
- Free SberPrime+ subscription, discounts on products from partner companies
- Referral bonus for recommending friends to join Sber's team.
