Company: Sberbank PJSC
Location: Moscow, Russia
Employment Type: Full-time
Prompt and AI Agent Engineer for automating banking support for legal entity contracts. Development and optimization of prompts for LLM agents, creation of multi-agent systems and quality assessment systems.
Tasks:
- Design, develop, and optimize prompts for complex LLM agents (planning, tool calling, memory management)
- Create interaction scenarios between multiple agents (Multi-Agent Systems)
- Develop systems for assessing the quality of agent responses: test datasets, metrics for accuracy, completeness, and hallucinations
- Conduct experiments for model selection and hyperparameter tuning
- Ensure security: design protective prompt systems, filter unwanted output
Requirements:
- Minimum 2 years of experience in a similar role
- Experience writing Python code: automating experiments, working with LLM APIs, creating evaluation scripts
- Understanding of LLM principles and prompt engineering: Few-shot, Chain-of-Thought, ReAct, Self-Reflection
- Practical experience building metrics and evaluation pipelines
- Experimentation skills: comparing models, varying hyperparameters
- Understanding of LLM security: prompt injection, jailbreaking, defense methods
Conditions:
- Hybrid work schedule
- Modern office at Kutuzovsky Prospekt, 32 with parking, a cafeteria, coffee shops, recreation areas, and a gym
- Annual salary review and annual bonus
- Over 400 educational programs from SberUniversity
- Adaptation program and mentor support
- Extended voluntary medical insurance, preferential family insurance, and corporate pension program
- Preferential mortgage and lending from Sber
- Free SberPrime+ subscription, discounts on products from partner companies
Skills:
- Python
- LLM
- Prompt Engineering
- Few-shot
- Chain-of-Thought
- ReAct
- Self-Reflection
- Multi-Agent Systems
- LangChain
- LangGraph
- CrewAI
- AutoGen
- RAG
- GPT-4
- Claude
- Gemini
- Open-source LLM
- Eval pipelines
- Prompt injection
- Jailbreaking
