I run an AI-native performance-marketing agency and build the agent fleet that runs it.

$ whoami

• GM, Pedal Performance — paid media for SMBs, delivered by a small team plus agents
• Operations, Pearmill — growth consulting

$ what I build

• Agent systems: Claude Code skills, hooks and MCP servers that encode judgment — when to act, what a check can't catch
• Data pipelines: ad-platform ETL → dbt → Supabase, reconciled to the P&L
• Client portals: reporting, boards and agent work-orders in one place

$ how I work

Every lesson becomes a rule, and every rule gets a check. Skills come out of real sessions, not demos, and anything destructive verifies before it deletes.

$ ls public/

• agent-tools — my Claude Code plugin marketplace; new skills and routines land here weekly

$ ping

New Paltz, NY