Building with AI, from agentic pipelines to art-directed systems.
I don't just design interfaces for AI—I use agentic coding, MCP architectures, and custom LLM pipelines to build and ship working software. Here are the tools, simulations, and systems I’ve built outside of enterprise walls.
AI Systems & Builds
Working software, desktop runtimes, autonomous agents, and simulations shipped outside of enterprise walls.
AI-Native Portfolio Architecture, Visual System & Role-Fit Engine
How I architected, art-directed, and vibe-coded this entire site as an AI-native product—combining a custom LLM job-fit evaluation pipeline and telemetry engine with a coherent, AI-generated isometric risograph illustration system.
Momentum: 3D Botanical Simulation & Terrarium Lab
A slow-paced 3D botanical terrarium and cellular simulation running at 60 FPS on Android and 90 FPS on Meta Quest VR—built with topological DAG hydraulics, Marching Cubes stone sculpting, and custom GLSL velvet shaders.
OmniMCP: Native Systems & AI Infrastructure
A high-performance Windows native desktop runtime, multi-client proxy, and dynamic context router for Anthropic's Model Context Protocol (MCP)—slashing startup token bloat by >85% and bridging OpenAPI endpoints without glue code.
VibE17: MeetupOS & Community Memory
An intimate, privacy-first community operating system for Walthamstow digital creatives—replacing impersonal ticketing forms with conversational progressive disclosure via AI concierge Morris, paired with real-time organiser Kanban and Gemini 3.8 Flash peer matchmaking.
PS: It's pronounced "blez" and it's he/him.
