I build the boring-hard parts around AI agents — the runtime, the guardrails, the landing zone — then I ship a poker table for friends.
- Building OpenShield — one command to run OpenClaw on Docker or Kubernetes with a security hook layer (
npx openshield@latest init) - Designing AWS foundations for agentic systems — multi-account landing zones, Well-Architected reviews, voice/autonomous agent platforms, sovereign patterns
- Comfortable in TypeScript, Python, Terraform, Docker, Kubernetes, AWS
- Shipping play-money Texas Hold'em with a live equity HUD
- Tinkering on XRPL / Xumm SDKs
- GitHub since 2017. Hireable.
Security is a runtime property, not a slide.
|
Security layer for AI agent runtimes One command to run OpenClaw on Docker or Kubernetes, with exec-policy presets, security hooks, and local JSONL audit logs. No dashboard tax.
|
Reference architectures for agentic systems Operational reviews for long-running agents, voice interfaces, secure tool use, and RAG at scale — judged by real load, cost, and incidents, not diagrams.
|
|
Multi-account foundation for agentic workloads OU structure, SCPs, PrivateLink, and cost attribution that survive agent delegation. Research vs production vs data planes without two separate foundations.
|
Play-money Texas Hold'em for friends Private rooms, AI seat fill, live equity HUD (win / tie / lose, outs, pot odds). No casino chrome. No real-money gambling.
|
Also in the same AWS set: aws-agent-platform · aws-sovereign-infrastructure · pizzabuilder-js
┌──────────────────────────────────────────────┐
│ CLASS Agent operator / infra │
│ SPAWN 2017 │
│ VEHICLE whatever gets the deploy there │
│ HP caffeine │
│ POLICY cautious (yolo on Fridays) │
│ QUEST ship things that stay up │
└──────────────────────────────────────────────┘
Keep shipping. Keep the runtime honest. Keep the table friendly.