AI assistants & data tools
I design and build AI chat assistants, multi-agent workflows, and linked-data research tools grounded in real, cited sources, with human-in-the-loop governance built in.

Software Engineer · Linked Data & AI Systems
The work spans Linked Art and CIDOC-CRM data pipelines, governed multi-agent AI, and full-stack systems from Cloudflare edge to pgvector. Seven projects are live in production today, from an 11-agent AG2 council to a museum workbench covering 13 collection APIs, each with a complete technical write-up.
From a simple site for your shop to custom software, I bring the same care to each one. Most projects start in one of these four places.
I design and build AI chat assistants, multi-agent workflows, and linked-data research tools grounded in real, cited sources, with human-in-the-loop governance built in.
I build booking systems, client portals, internal tools, and data dashboards on modern, reliable frameworks that grow with your business.
Sell online with secure checkout, real product photography, and inventory that stays in sync. It's the same setup behind the Sun & Rain Works storefront below.
Fast, mobile-friendly websites with your menu, hours, photos, and a map, plus a clear way for customers to call, book, or order. Easy for you to update day to day.
Not sure where your idea fits? Tell me what you need and I'll point you the right way.
Start a projectSeven shipped builds, every one live on its own production deployment. Each card links to a full technical write-up, source is on GitHub where public, and available for review on request where proprietary.
An AI council you can actually govern.
Autonomous multi-agent operations platform: a Next.js command hub driving an 11-agent AG2 council over an SSE bridge, governed by AIDD + TDD discipline, 3-tier RBAC, OpenTelemetry traces, and human-in-the-loop merge boundaries.
www.metagenauto.comApex domainSocrates anchors a hub-and-spoke council: four farm leads and five specialist lenses each weigh in from a distinct angle, with Glaucon steelmanning the rejected option. The operator streams in over an SSE bridge; every deliberation ends at a human gate that files a real ticket, and the AI never merges its own code.
Evidence-first exploration for Wikidata graphs.
A linked-data research app for searching Wikidata, inspecting entity evidence, and visualizing relationships, with optional AG2-backed agents that assemble a trustworthy, cited research picture.
www.wikidataexplorer.comApex domainOne graph for creators, works, sources, and cited interpretation.
A Next.js application for exploring creators, works, Wikidata relationships, and Linked Art records, backed by citation-grounded AI dossiers built on a typed data layer. Deployed at the metahistorybook.com apex, the same domain whose daily. subdomain serves Artwork of the Day.
www.metahistorybook.comApex · hosts daily.Source-backed museum data, gated to production by evidence.
A public-facing Linked Art workbench that turns museum collection APIs into source-backed exploration, rights-aware review, and evidence-gated publication workflows.
www.metamuseum.orgApex domainTurns Wikidata into a daily, cited art-history experience.
Each day it surfaces a painting by an artist born on today's date, then builds a deterministic, fully-cited article about the work and the people, places, and movements around it, live from open data. Deployed at daily.metahistorybook.com, the daily. subdomain of the MetaHistoryBook apex.
daily.metahistorybook.comSubdomainOpen-data Hawaii discovery, rendered as a fast public guide.
Engineered as an Astro static site for Hawaii travel, history, and culture, assembled from committed data files, reviewed public-source records, and deployed to Cloudflare Workers.
www.bigalohaguide.comApex domainA headless Shopify storefront, built end to end.
Built as a headless commerce storefront on Next.js and Vercel, integrating Shopify and Printful with custom AI product photography in place of stock mockups.
www.sunandrainworks.comApex domainLessons and patterns from building evidence-grounded systems, drawn from production work across all seven projects.
Every project here implements the same pattern: agents recommend, humans approve, nothing publishes silently. This is the architecture behind that discipline, illustrated across four different production systems.
Read →Meta Museum rates itself 8/10 and publishes the gap list. This article explains why distinguishing local test confidence from production proof is itself an engineering discipline, and what the criteria actually are.
Read →The gap between a correct AI answer and a traceable one is an architecture decision. This article examines how citation gates, claim-level verification, and persistent provenance records work across four production systems.
Read →A pipeline designed to keep moving is a pipeline designed to silently approve. This article examines how every agent in this portfolio has a named failure behavior, defined before it runs, tested like any other behavior.
Read →The technologies behind the production work above.
Interested in the work or want to talk through a project? Head to the contact page to reach me by email or on GitHub.