Y A R O S L A V V A S Y L E N K O

Independent researcher · Computational neuroscience · AI systems architecture

Poltava region, Ukraine · between forest and field

status: NFI  live · 4 subsystems · γ = 1.043 (McGuirl 2020) · G6 validated

I build systems that observe themselves.

Not in a metaphorical sense — literally. Each system I build includes a formal proof of its own state, a contract it cannot violate, and a diagnostic signal that says whether it's healthy or not. If it can't prove what it is, I don't trust it.

My background is unusual: competitive boxing, oil rig engineering, 7+ years of self-directed neuroscience. No CS degree. Everything I know I built from first principles — reading Sapolsky and Dubynin, running simulations at 2am, breaking things and understanding why.

I use Adversarial Orchestration — a methodology where every output passes through Creator → Critic → Auditor → Verifier before I trust it. I run parallel LLM agents, synthesize their outputs manually, and act as the human integration layer. The only results that survive are the ones that pass all four stages.

I don't prototype. I build systems with formal contracts, 99%+ test coverage, and evidence bundles that reproduce exactly on any machine.

Prompt engineering       →  canonical prompts with formal gates and honesty labels
Multi-agent systems      →  adversarial orchestration pipelines
Biophysical simulation   →  reaction-diffusion, fractal, topological dynamics
Scientific validation    →  Cohen's d, bootstrap CI, permutation tests
AI architecture          →  fail-closed, evidence-first, self-calibrating
Trading systems          →  geometric signal analysis, Kuramoto synchronization
System integration       →  formal contracts, adapter protocols, closed loops

γ_bio    = 1.043   zebrafish morphogenesis     (McGuirl 2020 PNAS)
γ_MFN    = 0.865   reaction-diffusion field    (Gray-Scott)
γ_market = 1.081   Kuramoto synchronization    (mvstack)
γ_DNCA   = 2.185   cognitive competition       (6 NMO operators)
divergence (bio, MFN, market) = 0.216 → UNIFIED

Three substrates converge on γ ≈ 1.0. DNCA operates at γ ≈ 2.0. Competition sweep reveals: γ minimizes to 0.756 at optimal competition. γ ≈ 1.0 is the topological signature of metastability itself.

▸ neuron7xLab/NFI             unified cognitive orchestrator
▸ neuron7xLab/MFN-plus        morphogenetic field network
▸ neuron7xLab/BN-Syn          biophysical neural dynamics
▸ neuron7xLab/CA1-LAM         hippocampal memory model
▸ neuron7xLab/ML-SDM          adaptive LLM behaviour