AI Engineer with 4+ years of experience building production LLM and ML systems. I focus on AI agents, RAG and Text-to-SQL, using Python to turn business workflows into working products.
| System | What I built | Result |
|---|---|---|
| Enterprise AI assistant Grand Line |
MCP tools for knowledge, service requests and tasks. Natural-language SQL reports, SSO, role-based access, approval before writes, task-aware model routing. | Request submission 10 → 3 min Report prep 60 → 10 min Inference cost −46% (benchmarks) |
| Manufacturing RAG Grand Line |
Page-level retrieval over scanned PDFs and engineering drawings, answers in Telegram. | 21,000 pages covered Lookup time −75% |
| Text-to-SQL agent RDP.RU |
Self-hosted LangGraph agent over ClickHouse, PostgreSQL and Hive with AST, RBAC and EXPLAIN guardrails. | Time to analytics 60 → 2 min Execution accuracy 0.82 50% of departments use it |
| Knowledge assistant RDP.RU |
Agentic RAG over 2,850 internal documents with retrieval and hallucination evals. | Recall@5 0.85 Hallucinations 2.4% HR load −40% |
| Threat Intelligence reports RDP.RU |
LLM reports on domains and IPs over async data collection and normalization. | Prep time 90 → 10 min 78% accepted without edits |
| Traffic classification RDP.RU |
CatBoost with Flink feature processing, inference and quality monitoring. | Accuracy 0.91 Macro-F1 0.88 Report cycle 3 days → 1 |
