I build AI
for people who
do the actual work.
Director of Revenue Operations turned AI practitioner. Senior PM background.
Thought leader. Practitioner. Multi-agent systems running on the desk under my monitor. A SaaS in the wild. The notes from doing the work, in public.
What I'm building
PurpleChipmonk
An 18-skill agentic operating system for the modern job search. Eighteen specialized AI skills, persistent state across sessions, and a self-improving playbook that learns from every correction. Not a chatbot. An actual operating system. Co-founded with Zach Berenzweig.
OpenClaw HQ
A three-tier multi-agent workforce running on my desk. Native Ollama on Apple Silicon hits ~18 tok/s on a 70B model. BullMQ + Redis orchestration. PostgreSQL + pgvector for skill memory. The whole thing runs while I sleep.
fowler-brain
A version-controlled AI operating system. Every AI I talk to — Claude, Cursor, local models — bootstraps from one source of truth. Lessons graduate to rules. Drift is reviewed weekly. The brain compounds.
Salesforce + AI Delivery
An eight-agent autonomous pipeline for Salesforce metadata generation. Apex, LWC, test classes, deployments — drafted overnight, reviewed in the morning. The Salesforce admin role, reimagined for an operator who refuses to do work twice.
Hire me
I build production AI for operators. Quality gates between LLM and user. Salesforce-fluent. Multi-agent shipping. Founding-engineer scope or focused consulting.
Six practice areas. Five live systems. One résumé that walks you through all of it. Visually-stunning PDF for the lazy reader; full HTML for the curious one.
What I write about
Local-first AI for operators
Running serious models on serious hardware that you actually own. Why the cloud isn't always the answer for people who care about cost, latency, and never sending their work somewhere they can't see it.
Multi-agent systems, in practice
Not the hype version — the version where you actually have to maintain them. What works when one agent has to hand off to another, what breaks, and why orchestration is harder than the demo videos suggest.
The Salesforce admin's AI playbook
Specific patterns for using AI inside the most boring-looking but most-used software on the planet. Apex generation, LWC scaffolding, test class authoring, debug log triage. The unsexy work, made faster.
Building in public, honestly
A 49-year-old solo builder shipping a SaaS while keeping a day job. The 90-day gates. The portfolio with a floor. The mistakes I'm making in real time. No pretending the path is straight.
A position
Most AI writing is for people who write about AI. That's not me.
I write for the people who do the work — the Salesforce admin, the RevOps lead, the solutions engineer, the SaaS operator — who read the AI hype, know it matters, and have no idea where to start.
I'm 18 months ahead of them. That gap is the whole point.
No frameworks I haven't shipped. No advice I haven't followed. No claims of expertise I can't demonstrate at a terminal in front of you.