Stanislav Melnikov

AI Engineer specializing in applied AI, multi-agent orchestration, production RAG, LLM evaluation infrastructure - with deep insurtech and fintech domain expertise, where AI reliability is a business requirement.
I build production-grade AI systems designed for reliability, observability, and graceful failure.
What I've shipped:
· Cut LLM agent execution 3x (180s → 60s) and first-token latency 2.7x (4s → 1.5s) - LangGraph node parallelization and caching in a live insurance broker multi-agent system
· Built a production RAG pipeline over unstructured insurance knowledge bases for claims support, end to end: chunking strategy, Qdrant vector database retrieval, structured outputs, tool-calling, human-in-the-loop fallbacks
· Deployed a multi-agent automation stack with MCP integrations (Slack, Gmail, Jira) that automated recurring operational workflows for internal teams
· Independently built and launched an AI novel co-authoring platform: multi-agent system (Python LangChain), RAG, faithfulness evaluation, Playwright E2E tests via AI agent - 89% E2E test coverage, live users

My engineering approach:
· Evaluation-first. LLM-as-judge loops, LangSmith trace observability, retrieval quality monitoring - standard practice, not something bolted on after a production fire.
· Domain depth past the API layer. Built an insurance broker SaaS with 20+ carrier integrations, understand the business workflows, and can turn ambiguous requirements into AI that solves the real problem - not just a technically correct version of it.
· End-to-end ownership. Root cause → architecture → eval pipeline → deployment → monitoring.

Core Stack:
LangGraph · LangChain · RAG Architecture · Vector Databases (Qdrant, pgvector) · LangSmith · Langfuse · Anthropic (Claude) · OpenAI · MCP (Model Context Protocol) · Multi-Agent Orchestration · LLM-as-Judge · Faithfulness evaluation · Human-in-the-Loop · AI Automation · Python · Docker · n8n · PostgreSQL · REST/SOAP/GraphQL APIs · Playwright · CI/CD

Open to AI Engineer roles where agent architecture, evaluation infrastructure, and production reliability are first-class engineering concerns.