Company:** Sberbank PJSC Location: Saint Petersburg Employment: 20-40 hours per week Salary: 40,000 – 80,000 RUB per month

Internship in a Sber team focused on developing intelligent risk analysis systems based on LLM and ML. The intern will participate in creating AI agents, training models, and developing PoC/MVP.

Tasks:

• Development of PoC and MVP AI agents based on LLM using langchain/langgraph frameworks
• Adaptation and training of Sber's LLMs using internal and external data (Prompt Tuning, LoRA, SFT, GRPO)
• Application of RL approach to improve the model's reasoning capabilities

Requirements:

• Bachelor's or Master's student in technical fields
• Knowledge of statistics and ML/DL algorithms
• Knowledge of Python and ML stack (PyTorch, Scikit-learn, Pandas)
• Understanding of modern NLP (tokenization, data processing, embeddings, pre-train models)
• Experience in designing and training neural network models (from CNN/RNN to transformers)
• Experience in building RAG (classic RAG, GraphRAG, Agentic RAG)
• Knowledge of classical algorithms and data structures
• Ability to apply threading, asyncio, multiprocessing modules
• Knowledge of OOP principles, design patterns, ability to write clean code, experience with Git

Conditions:

• Internship duration: 3-6 months, 20-40 hours per week, flexible schedule from Monday to Friday
• Work format: office-based, Saint Petersburg, Sber IT Hub, Ural'skaya St., 1
• Access to an extensive catalog of educational programs from SberUniversity
• Free subscription to SberPrime+ service and discounts on partner products
• Opportunity to receive rewards for referring friends