Winston Koh Illustration

About

Strategic Systems Architect & Full-stack Engineer

The Architect

I'm an INTJ (Architect) with a Strategist-first orientation. I treat challenges as systems to be understood — diagnose the physics first, then design around them.

My approach: identify the constraints of the arena, design a system that navigates them, and execute with precision.

Background

Education

15 years experience. BBM (Finance & Corp Comms), SMU 2011.

Credentials

CFA Level 1, ChFC®/S, CLU®/S, CFP® Modules 1-6, ACTA

Evolution

Finance → Insurance → Curriculum Development → Systems Architecture

2026

Ghost Brand Ecosystems, Zero-Point Architecture, Anti-Slop Protocol

Philosophy: The Anti-Slop Protocol

I navigate reality as it IS, not as it SHOULD BE. In an era of AI-generated noise ("Slop"), signaling requires friction.

This isn't cynicism — it's operational clarity. Moralism is a luxury; physics is mandatory. The question isn't "Is this fair?" but "What are the actual forces at play, and how do I position myself to survive them?"

"Navigate, don't litigate."

The Bionic Operating Model

I operate as a bionic unit — human cognition augmented by AI at every layer.

Human Intent (Top 1% Nuance) + AI Scale (Bottom 99% Execution) = 1000% Leverage

This follows the Architectural Isomorphism principle: The User provides the Meta-Cognition (How to think), and the AI provides the Execution (What to do).

The AI isn't a replacement; it's a force multiplier. I provide strategic direction and high-nuance judgment. The AI handles research, execution, and documentation at scale.

AI Proficiency

I work with state-of-the-art frontier models daily — not as a casual user, but as a practitioner who understands prompt engineering, agentic workflows, and how to orchestrate AI systems for production-grade output.

Frontier Models (2026)

  • Google: Gemini 3 Pro, Flash 3
  • xAI: Grok 4.1 (Thinking)
  • Anthropic: Claude Opus 4.5 (Thinking)
  • OpenAI: GPT-5.2, GPT-5.1-High

AI Capabilities

  • Prompt Engineering: System prompts, few-shot, chain-of-thought
  • Agentic Development: Multi-step autonomous workflows
  • RAG Systems: Vector search, semantic retrieval, knowledge graphs
  • AI Image/Video: Midjourney, DALL-E, Runway, Veo 2

Every project I ship is AI-native — from research and design to code and content. This isn't "using ChatGPT"; it's running a bionic production pipeline.

The Stack

Preferred Tools

  • Agentic IDE: Google Antigravity, Cursor, Windsurf
  • Reasoning: Claude Opus 4.5, Gemini Pro 3, o3
  • Backend: Python, Supabase (pgvector), Astro
  • Architecture: Flat-file Markdown, Protocol-Driven

Hard Passes

  • Black-box RAG: If I can't audit the retrieval, I don't use it.
  • Bloated Frameworks: No LangChain/AutoGPT. Primitives only.
  • Platform-Locked Memory: My context stays in Markdown.

Current Project: Athena (v8.2-Stable)

Athena is an AI-Powered Digital Personal Assistant operating on the Zero-Point Codex architecture. It enforces the Triple-Lock Protocol (Search → Save → Speak) to ensure every output is grounded in context.

It's not a chatbot. It's a living architecture of protocols, frameworks, case studies, and calibrated insights that compounds over time. Every conversation becomes intellectual capital.

View the public repository →

Want to work together?

Internal AI systems. Ops automation. Knowledge engines.

Get in touch →