Agentic AI systems, proven live.
AI Product Owner and agentic-systems builder. I ship multi-agent systems where quality is enforced, not assumed — every claim backed by an artifact you can open.
The system at a glance
One integrated stack: a versioned catalog of expertise, orchestrated into workflows, and executed by a runtime that gates quality at every step.
flowchart LR
A["Catalog<br/>38 agents · 37 skill folders"] --> B["Workflows<br/>10 BPMN pipelines"]
B --> C["Agentic Runtime<br/>eval-gate at every step"]
C --> D["Live verdict<br/>pass / fail, per step"]flowchart TB
A["Catalog<br/>38 agents · 37 skill folders"] --> B["Workflows<br/>10 BPMN pipelines"]
B --> C["Agentic Runtime<br/>eval-gate at every step"]
C --> D["Live verdict<br/>pass / fail, per step"]Proof at a glance
Agentic Runtime v0.6.0 — three workflows proven live, end to end,
against catalog v3.27.0:
- ✓ WF-001 — 3 / 3 steps passed
- ✓ WF-002 — 5 / 5 steps passed
- ✓ WF-003 — 7 / 7 steps passed
Per-step verdicts captured as JSON in the repository. See Live proofs.
Explore
-
Catalog
38 specialized agents and 37 skill folders, indexed in a CI-validated sidecar and spanning enterprise frameworks (Scrum, SAFe, ISTQB, TOGAF).
-
Workflows
10 orchestrated BPMN pipelines, from AI product scoping to project post-mortem — three already proven live.
-
Agentic Runtime
A headless TypeScript engine that runs the catalog with eval-gates enforced at every step. Proven live, end to end.
-
RAWLY
An evidence-first nutrition app where content is code. An audit exposed 4 hallucinated PMIDs in its legacy articles; an isolated fact-checker now re-resolves every identifier.
-
About
AI Product Owner / Agile delivery lead — 20 years, freelance. Business analysis (MOA) and web integration end to end.
Catalog currently at v4.0.1 · Agentic Runtime at v0.6.0 — both versioned,
with sources in their repositories.