AI Systems Architect · Platform Engineer · Creative Coder

Building at the intersection of scale, security, and intelligence.

I've spent 20+ years shipping software that has to work: BGP Anycast scrubbing networks that clean attack traffic before it hits the ISP, real-time DDoS mitigation at Akamai/Prolexic, HIPAA-compliant healthcare platforms, high-throughput data pipelines processing terabytes of claims data. The kind of systems where "it's down" means someone's getting paged at 3am.

Now I'm channeling that reliability obsession into AI. Multi-agent orchestration, computer vision pipelines, LLM-powered coaching systems, all built with the same rigor I'd apply to infrastructure that can't fail. Production AI isn't just about models; it's about observability, graceful degradation, and knowing when to put a human in the loop.

When I'm not architecting distributed systems or AI/ML projects, I'm probably upgrading things to work on Blackwell, writing shaders, working on game engines, or making websites do things they probably shouldn't.

Virtual Coach | AI/ML Systems Architect Computer vision pipeline for youth sports analytics. Pose estimation from monocular video, biomechanical metric extraction, and an LLM coaching engine that turns movement data into personalized feedback. PyTorch, CUDA, the whole stack.

JAZU Labs | CTO & Founder AI-native growth platform with multi-agent LLM orchestration, real-time generative content, and adaptive execution pipelines. The system grades its own outputs and refines its behavior across runs.

Fractional Engineering Leadership Advising early-stage founders on architecture, hiring, and scaling eng teams from chaos to structure without killing velocity.

• Production AI systems: multi-agent orchestration, real-time inference, self-healing pipelines
• Network infrastructure: BGP/Anycast routing, L3-L7 traffic engineering, DDoS scrubbing at scale
• Platform engineering: distributed systems, observability, SRE foundations, incident response
• Security architecture: bot detection engines, WAF, API hardening, compliance (HIPAA, SOC2, GDPR)
• Creative tech: WebGL, custom shaders, game engines, interactive experiences

EllyMUD - TypeScript MUD engine with MCP API for AI agent integration. My sandbox for testing multi-agent behaviors in a controlled environment. Play with friends, or simulate them with AI agents if you don't have any. (like I do)

Comic Orchestrator - Each character gets their own LLM brain. They react to scenes, write their own dialogue, and make decisions that push the story forward. Autonomous actors, not scripted puppets.

vibe-translate - CLI translation tool powered by Claude that actually sounds like a real person. Smart positional args, piped input, JSON mode for i18n files, 90+ languages.

ellyseum.me - My blog. Jekyll + Vite + TypeScript + Python (for generating RAG embeddings). WebGL particle systems, 3D card navigation, an AI chat widget with local LLMs running in the browser using WebGPU (with cloud fallback), and way too many shaders.

AI/ML: PyTorch · TorchVision · FlashAttention · CUDA · ONNX · LangChain Vision: OpenCV · MediaPipe · YOLOv8 · DeepSORT · MMPose Backend: TypeScript · Node.js · Python · PostgreSQL · Redis · GraphQL Infrastructure: Docker · Kubernetes · Terraform · GitHub Actions Frontend: React · WebGL · Three.js · GLSL · Vite Cloud: AWS · Azure · GCP · Akamai

_{"The future of AI-assisted development isn't a smarter single agent. It's a well-orchestrated team of focused agents that learn from every run."}