We are a team developing solutions based on machine learning and generative artificial intelligence within Sber's Corporate Investment Business (CIB).
Our goal is to create products that help Russia's largest companies and small businesses manage finances more effectively, automate routine tasks, and make investment decisions.
We are currently looking for talented individuals ready to dive into the development of real AI solutions for the complex B2B segment.
Responsibilities
- Conduct analysis of corporate data in collaboration with CIB analysts, participate in dataset labeling, and test hypotheses about business client behavior.
- Participate in cross-validation processes, A/B testing, and monitoring the quality of ML models in industrial operation.
- Develop and maintain content generation and document processing pipelines based on large language models (LLMs): from extracting entities from financial reports to summarizing business correspondence.
- Research and test new neural network algorithms and architectures for automating banking processes for legal entities.
- Create prototypes of intelligent systems (e.g., agents for supporting client managers or scoring models) using a modern MLOps stack.
Requirements
- Bachelor's or Master's degree in a technical field.
- Proficient in Python, writing clean code, understanding the principles of object-oriented programming (OOP) and SOLID.
- Practical experience in performing exploratory data analysis (EDA) and visualizing results.
- Understanding of the mathematical foundations of classical machine learning (regression, classification, clustering, quality metrics).
- Proficiency with tabular data preprocessing tools (pandas, numpy).
- Familiarity with the LLM ecosystem: experience with LangChain/LlamaIndex frameworks and the vLLM library for inference.
- Solved academic or pet projects using generative models (chatbots, RAG systems, working with embeddings).
Will be a plus:
- Experience with distributed computing (Apache Spark, PySpark) for processing large data volumes.
- Practical experience in creating AI agents that use external tools (tool calling) for performing actions.
- Experience using the API of the GigaChat family of models or other Russian foundational models.
- Skill in writing SQL queries for independently extracting data from storage.
- Basic knowledge of Docker for containerizing your applications.
Conditions
- Paid internship for 3 months with the possibility of subsequent full-time employment.
- Office-based format (40 hours per week), allowing for full immersion in team tasks, with flexible hours.
- Work at Sber's IT hub office located at: St. Petersburg, Uralskaya St., Building 1 (10 minutes on foot from Vasileostrovskaya or Sportivnaya metro stations).
- Real tasks under the guidance of experienced mentors – you will work on products used by thousands of entrepreneurs.
- Access to powerful computing resources (GPU clusters) and the bank's internal data services.
- Adaptation program, internal training in Deep Learning and NLP, as well as access to the corporate library.
