AI Developer (backend)

We are MStroy — residents of Skolkovo and INNOPOLIS, an actively developing, innovative IT company with roots from Tyumen, operating on the basis of a large bridge construction company in Russia, JSC "Mostostroy-11".

We are developing a digital construction platform to increase the efficiency of construction activities by enabling prompt decision-making based on up-to-date online analytics (on deadlines, personnel, supplies, finances).

Accredited by the Ministry of Digital Development of the Russian Federation (deferment is possible during partial mobilization).

To develop the MStroy platform, we are looking for an employee who is passionate about their work!

WHAT YOU WILL DO:

• Develop AI agents: design logic, write prompts, configure tool calls (tools) and integration with external APIs.
• Develop backend services in Python (FastAPI) using REST and GraphQL APIs, through which the agents operate.
• Design and optimize data schemas in PostgreSQL and ClickHouse for efficient work with large volumes of structured data.
• Deploy and configure open LLM models on the company's own infrastructure (vLLM, Ollama, etc.).
• Integrate agents with the MStroy platform and external systems.
• Evaluate and improve the quality of agent work on real data: implement metrics, logging, monitoring, and prompt versioning (Langfuse).
• Independently bring tasks from initiation to completion in a production environment.
• Write clean, documented, and testable code, adhering to best Python practices and team standards.

Our stack

• Backend: Python 3.9+, FastAPI, REST API, GraphQL
• Databases: PostgreSQL, ClickHouse
• Vector Databases: Qdrant, pgvector
• Agent Frameworks: LangGraph, LangChain (understanding principles is more important than knowledge of a specific framework)
• Model Deployment: vLLM, Ollama (local deployment of open models)
• Monitoring and Versioning: Langfuse
• Containerization: Docker
• Version Control System: Git / GitLab
• Operating Principle: Only open source, deployed on own infrastructure

WHO WE WANT TO SEE IN THE TEAM:

• Higher technical education (completed).
• Industrial Python development experience of 3+ years (backend services, APIs, database work, asynchronous programming).
• Deep understanding of agent systems architecture: orchestration, tool calling, state and error handling. Experience with LangGraph or similar.
• Ability to choose and justify an approach to data retrieval: exact search (SQL), full-text, regex, vector, hybrid. Understanding where vector search is overkill.
• Experience with local deployment of open models (vLLM, Ollama), understanding resource requirements and trade-offs between quality, speed, and cost.
• Skills in evaluating agent quality: measurement with metrics and tests on real data, logging, monitoring, prompt versioning.
• Experience in deploying solutions to production and supporting them under load (fault tolerance, degradation, error handling).
• Independence: ability to bring a task from initiation to completion without constant supervision.
• Experience with Git (GitFlow, Merge Requests, Code Review).

Will be a plus:

• Experience with fine-tuning open models.
• Experience in construction or related fields.
• English language proficiency for reading technical documentation.

WHAT WE ARE READY TO OFFER:

• Official employment under the Labor Code of the Russian Federation from the first working day. Full social package: white salary, paid sick leave, vacations, all guarantees in full.
• Voluntary Medical Insurance (DMS) with extended coverage (including dental and preventive check-ups) — available after 6 months of employment.
• Internal and external training system. We compensate for specialized courses, conference participation, and regularly conduct internal workshops.
• Assistance in obtaining an IT mortgage on preferential terms.
• A cohesive team with live communication and support.
• For local employees, only office-based work is available in Tyumen, Smolenskaya St. 41 (with its own parking).

Candidates will be required to complete test assignments as part of the selection process.