Viktor Skobeliev

I’m an AI Automation Engineer with 6+ years in IT and 2.5 years building AI automation that runs in production for real businesses—LLM-powered workflows, agent systems, and API integrations. My core principle is simple: the LLM handles judgment and language, and everything that must be exact moves into deterministic, tested code.

I’ve rebuilt production systems by replacing drifting LLM output with tested calculators and generators—turning “AI demos” into systems a business can depend on. I focus on deterministic AI architecture to remove run-to-run drift, plus rigorous testing for non-deterministic systems using regression harnesses, simulators that mock external APIs, and reference batteries with hand-computed expected values.

For end-to-end ownership, I build from business problem → architecture → build → deploy → documentation → handover. I’m currently the sole engineer of my own SaaS product, covering everything from the image-processing engine to billing, and in my professional work I’ve delivered automations like Figma-to-HTML production pipelines, estimation agents with structured JSON decisions, and CRM/payment-verification workflows that escalate exceptions to humans instead of guessing.