Artem

Romanov

I build AI automation that does real operational work, then verify it actually did the work right. Day job: marketing manager at an HR agency, where I wrote the software that runs our hiring: 6 Avito accounts, 5,319 ads uploaded via API, a desktop app with 657 automated tests running 24/7 on a Windows Server VDS. What I do for clients: - AI agents and automation: Python 3.12, Claude Agent SDK, OpenRouter, n8n, MCP, Playwright, FastAPI, Telegram bots (aiogram, Telethon) - Data and content pipelines: parsing, Google Sheets / Apps Script reporting, LLM content with a human acceptance checklist - LLM output evaluation: the same task on several models, expected answer written first, then checked against a checklist - Marketing ops: Yandex.Direct full cycle, funnels, lead generation, analytics I work async and in writing (English written, Russian native). You get working code and a short report: a number goes into it only from a command that was actually run.