Work With Me
25 years building software teams. Now I build AI agent teams — and help others do the same.
What I Do
AI Operations Design
Design multi-agent workflows for your team. Agent roles, coordination layers, handoff protocols, error triage — the operational infrastructure that makes AI agents actually useful in production.
Fractional AI Advisory
Ongoing advisory for founders and engineering leaders integrating AI into their development workflow. Strategy, architecture review, and pattern recommendations based on what actually works.
Agent Team Architecture
Help you design your first (or next) AI agent team. Which roles to create, how to scope them, what coordination files to use, and when to delegate vs. build.
AI Product Strategy
For companies looking to add AI capabilities to existing products or build new AI-native products. Product strategy informed by 25 years of shipping software and running multi-agent systems daily.
Background
I've spent 25 years in tech — as a developer, architect, CTO, founder, and consultant. I've built teams from 2 to 50, shipped products used by millions, and learned (the hard way) what makes software organizations work.
Now I apply those operational design principles to AI agent systems. MonkeyRun is my live lab — a startup studio where I run 20+ AI agents across 6 projects, documenting every pattern, failure, and breakthrough.
The hard part of AI isn't the models. It's the operations. Orchestration, coordination, context management, delegation patterns, error recovery — these are operational design problems, and I've been solving operational design problems my entire career.
What Makes This Different
Battle-tested patterns, not theory
Everything I recommend comes from running multi-agent systems in production daily. 22 documented patterns and counting.
Operational design, not prompt engineering
Prompts are commoditizing. The lasting advantage is how you design the system around the agents — roles, coordination, monitoring, error recovery.
Honest about what doesn't work
Our COO crashes every 24 hours. Our PM writes specs longer than the code. I'll tell you what fails, not just what succeeds.
25 years of context
The patterns that work for AI agent teams are the same ones that work for human teams. I've been learning those patterns since before LLMs existed.