Vacancy Description
The team is engaged in the development and support of in-house systems.
Responsibilities
- Develop new AI agents and enhance existing ones using Python: dialogue scenarios, request routing, context and dialogue state management;
- Integrate LLMs via API, configure prompts, structured responses, and tool calls;
- Connect AI agents to corporate services and tools via REST API, MCP, and A2A;
- Develop and refine backend components of AI services using FastAPI or Flask;
- Implement agent skills, error handling, logging, and verification of action execution correctness;
- Test agent performance, improve response quality, integration speed, and stability;
- Integrate agents into a chatbot.
Qualifications
- Familiarity with FastAPI or Flask, HTTP/REST API, JSON, and asynchronous Python programming;
- Experience with SQLAlchemy;
- Experience interacting with LLMs via API; understanding of prompts, context windows, tokens, and structured output;
- Understanding of AI agent architecture principles, tool calling, and request routing; familiarity with the purpose of MCP and A2A;
- Basic knowledge of ML/NLP: embeddings, semantic search, relevance evaluation; experience with one of the vector databases (Qdrant, Weaviate, or similar);
- Understanding of how to evaluate LLM and agent quality: correctness and relevance of responses, quality of tool calls, latency; basic understanding of search metrics;
- Skills in working with Git, Linux terminal, and remote servers via SSH; ability to read logs and troubleshoot errors;
- Familiarity with Docker, Redis, and the basics of Python application testing.
- Will be an advantage
- Knowledge of essential security practices for LLM applications and AI agents: protection against prompt injection, secure handling of untrusted data, access control to tools, and prevention of confidential information leaks;
- Understanding of MLOps/LLMOps principles: logging and monitoring, model and prompt versioning, automation of AI service testing and deployment;
- Understanding of database schema migration principles and familiarity with Alembic for managing changes in models and table structures;
- Understanding of architectural approaches to building AI systems: when a simple LLM pipeline is sufficient, and when an agent-based approach is justified; application of caching, asynchronous processing, and various orchestration strategies considering quality, performance, and reliability requirements;
- Understanding of LLM specifics and limitations, common problems with context handling, causes of hallucinations, and main approaches to improving response quality and stability;
- Familiarity with LangChain and LangGraph.
What we offer
- Work schedule of 40 hours per week;
- Professional development within teams of professionals and an experienced mentor ready to share knowledge;
- Mobile communication covered by the company;
- Special offers and discounts from MTS ecosystem products;
- Own corporate university, online library, Best Benefits (employee application with various discounts);
- Sports: online stretching, yoga, breathing exercises.
