From HP-UX and Solaris through Kubernetes, OpenShift, and modern AI runtime work.
IBM / eu-LISA, Poste Italiane, and TIM shape the reliability baseline.
Value-Shore platform work plus public Pi and Codex packages with private systems listed only.
Apple Silicon, macOS companions, and controlled multi-provider routing by default.
Selected systems
A few systems that show the actual shape of the work.
Clickable names are current public GitHub proof. Private systems stay named, not linked.
01
SmoothOperator / agents
Private multi-tenant AI orchestration platform with workflow controls, document layers, artifact delivery, tenant isolation, and multi-provider execution.
02
pi-multiagent
Model-native Pi package for isolated same-session multiagent delegation with governed roles, graph runs, status inspection, and result retrieval.
03
pi-continue
Continuation package for safe mid-run compaction, package-owned proof, and same-session task resumption without transcript replay.
04
cdx-claude
Codex plugin and MCP server for bounded Claude Code delegation through isolated jobs, status, tails, results, diffs, and cleanup.
05
the-seven
Privacy-first multi-model council that keeps drafts, critiques, rankings, run details, and final answers inspectable.
Public supporting:
pi-exa-tools
ollama-pub
homebrew-agenterm
Listed-only private systems:
cdx
er_guaina
cdx-relay
Veneer
local-doc-studio
voices
mac-local-image
Multisub
Career arc
Long path. Consistent pattern.
Operational truth first, architecture second, AI systems built on top of both.
1999 → 2007
Unix / Linux administrator
HP-UX, Solaris, Linux, clustering, hardening, backup, recovery, and learning to respect the substrate.
2008 → 2015
Systems engineer
Virtualization, storage, middleware, and turning infrastructure into something repeatable rather than improvised.
2016 → 2024
Solution / platform architect
Kubernetes, OpenShift, reliability, isolation, and platform boundaries shaped around real operational constraints.
2025 → now
AI systems builder
Agent runtimes, orchestration planes, multimodal tools, and local-first products built with operator-grade discipline.
Operating principles
Short rules that survive contact with production.
Reliability
Design for predictable degradation, not heroic rescue.
Operator clarity
Logs, controls, and contracts are part of the product surface.
Privacy-first
Keep data, models, and control close to the operator when the product allows it.
One canonical path
Competing implementations are drift, not optionality.
Evidence over hype
If a system cannot be explained and verified, it is not finished.
Contact
If the problem is messy, send the mess.
I work best on systems that need clarity, reliability, and a calm operating model.