👋 Hi, I'm Maksim
AI/Backend Engineer designing and shipping production-grade LLM agent systems — from architecture decisions to deployment. I care about reliability under real-world constraints: structured outputs, guardrails, graceful degradation, and systems that fail safely. Particular interest in applying this to fintech, real estate, and IT-service domains, where correctness and data trust matter most.
Focus areas:
- Multi-agent orchestration (LangGraph, LangChain) — coordinating tools, state, and failure recovery across long-running agent workflows
- RAG systems on Azure OpenAI / OpenAI with Weaviate, Qdrant, pgvector
- MCP-based tool integrations with retry/repair logic for production reliability
- FastAPI microservices — PostgreSQL, MongoDB, RabbitMQ, Docker
Selected work:
- invest-pulse-agent — Telegram AI product delivering investment portfolio digests and market outlook, with a second module for SaaS idea generation — built and shipped end-to-end as an independent product
- workspace-agent — ReAct agent on Claude autonomously orchestrating Slack, Jira, Calendar & Email through a typed FastAPI service; designed for multi-tool decision-making, not scripted flows
- support-triage-agent — production ticket-triage service with prompt-injection guards, structured Pydantic outputs, and automatic repair-prompt retries for safety at scale
- multi-agent-soccer-orchestration — LangGraph multi-agent system combining live data, web search and Azure OpenAI
- weather-rag-assistant — RAG microservice on Azure OpenAI, LangChain and Weaviate
Always open to interesting conversations — startups, ambitious products, or unconventional problems. Reach out on Reach out on LinkedIn or via email.