Egor Iakobson

At UniQee, I architect the LLM Arbiter engine, an autonomous agent that processes simultaneous blind inputs and manages dynamic state machines. I scaled its architecture for 4,000+ users, 200+ complex daily LLM queries, and 40+ AI personas.

I design XML prompt guardrails, JSON-data mappings, validation workflows, and Tone of Voice logic that turn complex psychological parameters into reliable multi-agent interactions. My work focuses on making language models follow strict rules and produce clean, machine-readable outputs with near-zero hallucination.

Previously at Next Genium and as an independent Interactive Logic Designer, I turned narrative chaos and unformatted data into variable-driven logic trees, decision matrices, procedural templates, and state-tracked interactive systems.

I also build practical AI products: a RAG writing coach for Telegram, a LegalTech CRM automation prototype, a booking tool with double-booking prevention, and a Telegram AI companion engine. I connect LLMs, vector databases, low-code tools, and external APIs to create functional business tools and autonomous agents.