NeoZorKResearch

About

Rostyslav Shcherbyna

Portrait of Rostyslav Shcherbyna

Rostyslav Shcherbyna is an independent researcher and engineer. He started as a system administrator, project manager and QA team lead, moved into trading on US equity markets, spent nine years in analytics as a co-partner at Filodema Analytics, and then three years as a quantitative researcher at SmartTrader.

He now works independently under NeoZorK Research across time series in many domains, econophysics, AI agents and local language-model infrastructure, verification and reproducibility, knowledge engines and data infrastructure. The approach rests on pre-registration, out-of-sample testing and the equal publication of negative results. The mission: help build innovation for humanity.

15+ years hands-on · ORCID 0009-0000-3773-5789

From hands-on to research direction

Path

Hands-on years

Building and operating systems myself: code, data, markets.

  • System administrator, project manager, QA team lead

    Operations, delivery and quality foundations in software teams.

  • Trader, NYSE and NASDAQ

    Thinkorswim and Bloomberg Terminal; the practical side of market risk.

  • Co-partner and analyst, Filodema Analytics

    Nine years of analytics, general work and conclusions.

  • Quantitative researcher, SmartTrader

    Three years of quantitative research.

Research direction, AI-assisted

Since 2024: I set the questions, design, specify and validate; AI tools implement.

  • Independent researcher and engineer, NeoZorK Research

    Time series across domains, econophysics, AI agents, verification and data infrastructure; pre-registered experiments, honest negative results.

  • Creator of Monte-Neo and ClaimBound Evidence

    Open-source verification tooling for strategies and public claims.

  • Canonis

    A local-first engine for evidence-bound knowledge, in development.

  • NeoArc (R&D)

    A verification-first alternative to slow LLM-only code generation.

Honesty

How the work is made

Since 2024 new code is written AI-assisted (Claude Code, Cursor, local language models). I set the questions, design the architecture and validate every result; the tools type faster than I do.