Multi-Agent Debate Arena
dinos.ext/battlefield
A multi-agent debate arena where Claude and Grok go head-to-head in real-time structured argumentation — and you step in as the third agent. Intervene, force a roleswap, or just watch the logic unfold. No judge, no verdict, no saved conclusions.
AI-powered video summaries & tags
Distilloon
An interactive AI-powered tool that distills YouTube videos into their core meaning — visualized as colorful balloons carrying keywords that float, compete, and pop. Watch AI scan, select, and crystallize a video's essence in real time.
Civic AI Platform
Montgomery Pathways
A dual-interface civic platform combining AI-powered policy simulation for city planners with conviction-aware career navigation for underserved residents — driving inclusive economic growth in Montgomery, Alabama. Build for real people, real problems, with live & raw data.
Molecular Exploration Platform
Black Core Logic Gate
Logic-driven multi-agent framework for molecular exploration in Physical Material Science. Bypasses LLM hallucinations via a Logic Gate architecture integrating RDKit for rigid physical validation. Designed to discover extreme entropy structures like Stealth and Void Black.
Honeypot Intelligence
Phantom Trace
Synthetic attack logs fed through an LLM pipeline for classification, threat scoring, and MBTI attacker profiling. Faker vs Markov generation methods compared, plus observations from two live sites. What attracts modern attackers?
Physical Artwork
Mathematical Art Portfolio
A decade-long practice designing and building mathematical art at institutional scale — where equations become objects people can touch. The foundation behind everything else here.
About
I build AI systems that sit at the intersection of language, logic, and design. My work spans LLM architecture, multi-agent orchestration, and the interfaces that make these systems intelligible to humans.
As a Technical Lead, I bridge the gap between research and product — translating complex model behavior into actionable engineering decisions, and shipping systems that people actually trust — built around how humans think, not just how models perform.