Work

What I can show is here. What I cannot show shaped it: 50+ agentic proof-of-value systems built for banks, insurers, health companies, and retailers stay behind their walls, but every pattern they taught me ends up in a field guide.

LLM Inventory

✓ shipped

A static read of a repo that lists every model it calls, whichever vendor they belong to, scores how hard the codebase would be to move, and maps each id onto the destinations you actually want to run on (Cursor, Anthropic, Google, xAI, Antigravity, Copilot, Replit, Bedrock, OpenRouter, or your own hardware). Model catalog as of September 2026. Zip, folder, and private-repo scans never leave your browser.

TypeScript Vercel Functions Astro run it →

AURUM · AG-UI wealth onboarding

✓ shipped

A private-bank onboarding agent that streams the whole interface, not just text. Generative UI cards, risk gauges, and human approval desks over one SSE stream, orchestrated by a process engine. Runs without API keys in shadow mode.

AG-UI Camunda 8 Claude Gemini Node.js React repo ↗ episode →

Tool Credit Score

✓ shipped

A deterministic linter that scores MCP tool specs 0 to 100 for agent-readiness and auto-fixes the failures. Born from eval findings that agents skip tools because of bad specs, not bad models. Five weighted categories, no model in the loop, and the whole thing runs in your browser.

AURUM Pays · AP2 agent payments

✓ shipped

One of the first working implementations of Google's AP2 v0.2 outside their own samples, wired to an external governance layer. A shopping agent buys FIFA tickets overnight under a cryptographically signed mandate it cannot override.

AP2 v0.2 Camunda 8 Claude SD-JWT DMN React repo ↗ episode →

Mortgage concierge · multi-agent A2A

✓ shipped

Multi-agent mortgage pre-qualification with a judge agent gating decisions at a confidence threshold, and a locally-hosted research agent behind a data-privacy boundary. Became the worked example in Camunda's official A2A blog series and a Gemini 3 hackathon listing.

A2A Gemini 3 Camunda 8 DMN repo ↗

Four-axis agent eval harness

✓ shipped

Built when nobody was asking for agent evals, now it is the first question in every room. Measures cost per case, accuracy against a gold set, tool-skip rates, and temperature sensitivity for process-embedded agents. Field-validated on live enterprise workloads across 20+ models. The method is public; the code is not.

Evals Claude Gemini Cohere Bedrock

Cohere connector for Camunda

✓ shipped

A marketplace-submitted connector bringing Cohere's models (chat, structured JSON, classification, embeddings, reranking) into process orchestration as element templates. Built with sovereign-data deployments in mind, where a Canadian model and a self-managed cluster matter.

Cohere Camunda 8 Element templates repo ↗

what-are-loops

✓ shipped

The anatomy of the agentic loop, the primitive every major coding tool independently converged on, published as a spec, a deep-dive, and a motion graphic built with the loop it describes.

Claude Code Remotion repo ↗ episode →
esc