Description
We are looking for an outstanding Data Scientist/ML Engineer to join the product team implementing and developing artificial intelligence in the EdTech sector at SBER.
You will be responsible for developing ML pipelines and AI systems using cutting-edge AI technologies: NLP, LLM, MLLM, RAG, AI agents, and multi-agent systems, specifically based on SBER's flagship model – GigaChat.
Your primary goal, alongside data analysts, prompt engineers, and developers, will be to create production systems capable of processing large volumes of multimodal data and delivering results aimed at effectively solving business problems.
You have a unique opportunity to participate in shaping cutting-edge AI native EdTech solutions as part of a team of strong and ambitious professionals focused on results.
Responsibilities
- Rapid hypothesis testing (Proof of Concept): data preprocessing, selection, testing, and fine-tuning of models and functionality (e.g., function calling), calculation and improvement of metrics necessary for solving the business task, EDA;
- Participation in the preparation of test and validation datasets, development of requirements for data, its format, quality, structure, volume, elaboration of appropriate labeling methodologies necessary for the high-quality operation of AI systems;
- Development, implementation, monitoring, and optimization of ML pipelines and AI systems, particularly RAG (LLM, cross-encoder, ranking models, etc.), including for multimodal data (text, video, audio);
- Joint analysis of functional requirements with the team, solution design, planning the optimal functional implementation architecture;
- Participation in developing optimal integration scenarios with data sources (including different formats and modalities) together with related teams, setting up optimal ETL processes considering business requirements.
Requirements
- Knowledge of Python and data processing libraries (pandas, numpy, matplotlib, seaborn, pytorch, sklearn, BeautifulSoup, etc.)
- Knowledge of classical and modern methods for working with textual data: tokenization, stemming, lemmatization, NER, sentiment analysis, clustering, TF-IDF, embeddings; transformers (BERT, etc.), LLM, MLLM, etc.
- Understanding the principles of models in the field of speech recognition/synthesis and image generation;
- Building LLM, RAG pipelines, agent and multi-agent systems using relevant tools: LangChain, GigaChain, LangGraph, function calling, vector databases, ETL, etc. Tracking trends and innovations, promptly testing new tools to solve applied business problems;
- Understanding MLOps principles, participation in processes for deploying and supporting developed AI solutions in a production environment;
- Experience/understanding of working with technologies: Docker, fastAPI, gRPC, REST, etc.
- Proficient use of AI for analysis, generation, and automation.
Will be a plus:
- Understanding the nuances of prompt engineering and relevant experience;
- Work experience/knowledge of Java;
- Understanding the principles of classical ML, recommender systems, CV.
Conditions
- Hybrid work format (modern office in Moscow on Kutuzovsky Prospekt);
- Preferential mortgage lending conditions;
- Free SberPrime+ subscription, discounts on products from partner companies: Okko, Sber Market, Mega Market, Samokat, Eapteka, and others;
- VHI from day one and insurance discounts for family members;
- Corporate pension program;
- Children's recreation and gifts covered by the Company;
- Company-sponsored training: online courses, unlimited access to the library, and training at the Corporate University, workshops, meetups, and the opportunity to obtain new qualifications;
