For years I had ideas I couldn't build without resources I either didn't have, or had to convince others of the need and wait on someone's calendar. AI changed that.

Browser-based system that converts unstructured written instructions into structured 2D symbol charts. My one real product, currently heading to alpha.

Built around an anchor-placement data model, constrained integer-grid layout, and composable templates for complex nested structures. Solves an end-to-end parsing problem no existing tool handles. Expanding into a suite of related tools.

The project I built to actually understand AI and learn its capabilities. An AI-native system for work: submit intent in plain language, it decomposes the goal into a task graph, routes work to specialized agents, and executes under human governance.

My deep end of the pool. Building it forced me through orchestration, durable workflows, agent design, knowledge persistence, and governance, the systems concepts that make AI genuinely useful instead of just impressive.

TypeScript · Temporal · PostgreSQL/pgvector · Next.js

AI system governing content generation and voice consistency across a 480-document creative canon for an original world.

Enforced tonal registers per character type; a repeatable write → critique → revise → human-gate loop designed to produce consistent, canon-accurate output at scale without drifting into generic AI prose.

Multi-stage audio segmentation and voice transformation pipeline.

Segments audio by character and transforms voices using ElevenLabs AI. A fast prototype, designed, built, and working in a single session, demonstrating multi-step AI API integration.