AI Systems Architect building trading systems, multi-agent infrastructure,
and local LLM pipelines that run without you.
Not forks. Not wrappers. Designed and engineered from scratch.
AI-native 3-lobe agent brain. Concept web with semantic decay, memory promotion, and context injection at boot. The reasoning core behind everything I run.
TurboQuant (ICLR 2026) for Node.js/Bun via napi-rs. MLX KV cache for Qwen3 on Apple Silicon — 32k → 128k context. 7 research findings, 3 production bugs fixed.
Autonomous hedge fund architecture. Nova as CEO, Ministral as private reasoning core, shadow-testing pipeline, always-online agents with auto-restart on failure.
QP-optimal multi-strategy capital allocator. EWMA covariance, adaptive gamma, OSQP solver, live bridge, and Telegram alerts. Risk flags: OK → DE_RISK → KILL.
Polymarket 5-min BTC binary scanner. Gaussian oracle edge detection with fractional-Kelly sizing, CLOB execution, and automated market filtering.
Nova routes. Theorex remembers. Agents execute.
const LORD_ZYTHOZ = {
location: "Sydney, Australia 🇦🇺",
role: "AI Systems Architect",
focus: ["Multi-agent orchestration",
"Local LLM inference",
"Quant trading"],
philosophy: "Sovereign infra. My models.
My rules. Agents that
earn their keep.",
};