Serhii V.

I build AI systems that run on their own — and stay accountable for them like a CTO, not a contractor. Recent work, all in the portfolio below with numbers: a video pipeline where a script goes in and a finished ad comes out (8 LangGraph agents, 30 parallel H100 renders, ~$2.50 per finished minute); a RAG support agent for a crypto↔fiat exchange rebuilt from 40% to 95.6% on a blind human audit, zero hallucinated services; a voice control plane on Vapi + FastAPI that reads code, runs commands and dispatches work to a coding agent — 285 tasks in production. WHAT I DO Agentic systems — multi-agent workflows (LangGraph), RAG with hybrid retrieval and output guardrails, evaluation you can trust before go-live AI video automation — WAN / LTX / FLUX on GPU workers, ElevenLabs voice locked to Whisper timings, Remotion + ffmpeg assembly, vision QA loops that re-render failed scenes Voice agents — Vapi / Twilio, tool-calling, call-flow and audio debugging on live traffic Fractional CTO — architecture and codebase audits, GPU/LLM cost reduction, taking over messy platforms, leading senior engineers HOW I WORK I tell you what will and won't work before you spend money. Short cycles, a working demo at every step. Everything ships with observability and a handover, so the system doesn't depend on me. Track record: $300K+ on Upwork, Top Rated, 100% Job Success, 8,800+ hours. 15+ years building production systems, founder of Genesis (AI video product). Stack: Python (FastAPI), Laravel, Node.js · OpenAI, Anthropic, DeepSeek, open-weight models · LangGraph, Qdrant, Postgres, Redis · AWS, GCP, RunPod · Vapi, Twilio, ElevenLabs, Remotion Good fit if you need a working AI system, not a prototype. Message me with what you're trying to automate — in one reply I'll tell you whether it's realistic, roughly what it costs to run, and how I'd build it.