Brendan / Selected work

Work, with honest labels.

A day job in production and a couple of independent projects. Everything is described in general terms; nothing uses non-public data.

01Projects

  1. 01

    Production AI at Arcadia

    In production

    Agentic assistants, LLM evaluation and observability, and document-workflow orchestration on real, regulated healthcare data.

    My role: VP of AI & Automation. I lead the work and build hands-on.

  2. 02

    Fact Riot

    In development

    A tool in development intended to turn links or posts into claim-by-claim receipts with sources. The aim is for checked claims to build a public track record for people and AI agents.

    My role: Creator. Independent, non-healthcare project.

  3. 03

    Auto-Tunez

    Experiment

    A personal music-interface experiment built with React and TypeScript. Exploring the connection between mood, music and visual atmosphere.

    My role: Creator. Independent project in development.

  4. 04

    Autarkic

    Experiment

    A draft specification for AI agents as accountable participants: discovery, reputation and dispute resolution on existing agent standards.

    My role: Author. Early design work, not a product.

02At Arcadia

Where the day job focuses.

01

Agentic AI in regulated healthcare

Multi-agent assistants doing real work under production guardrails: routing, retrieval and tool use where mistakes have consequences.

02

LLM evaluation & observability

Eval-driven LLM operations: tracing, scoring and datasets that keep production prompts honest over time.

03

Document-workflow orchestration

LLM document-processing pipelines: extraction, routing and human review at volume, in workflows that can't silently fail.