A private multi-agent AI system
An always-on assistant with a tag-based tool protocol, rolling long-term memory, and vision delegation — backed by a private Bedrock/Claude proxy I deployed and operate.
Full-stack systems engineer bridging applications, integrations, and applied research. I build the tooling, agents, and harnesses that turn hard problems into shipped products.
Ten years shipping mobile apps, developer tooling, and computational-math tools — lately building multi-agent AI systems and applied-optimization prototypes. I like closing the gap between a hard technical capability and the people who need to use it.
An always-on assistant with a tag-based tool protocol, rolling long-term memory, and vision delegation — backed by a private Bedrock/Claude proxy I deployed and operate.
A best-first (A*) variant of Knuth's Algorithm X that yields solutions cheapest-first under a cost function. Originally Python / NumPy / pandas; ported to TypeScript to run live in the browser.
Reducing exact cover to a form a neutral-atom quantum processor can optimize — a runnable end-to-end demonstration.
Extracts palettes by training a self-organizing map over an image on a chosen topology (torus, Klein bottle, and more), then reading the organized colors back out.
Interactive apps and integration builds shipped under my own studio — from real-time games to web tooling and third-party service integrations.