In 2008, I prototyped an ontology-driven knowledge management system using Endeca, metadata, and real data — a direct precursor to today’s RAG+LLM architectures. In 2025, SCRY applies similar concepts to State governance.

I demoed this project to a board of directors but it wasn't approved and died after my company lost their contracts. Manual procedures were creating inconsistency in plane repairs and this was an attempt at automated standardization. It's about twenty slides but I'm only showing a few with relevant design features.

✈️ 2008 vs 2025

Layer	2008 Knowledge Labs	2025 SCRY (AI Advisor)
Problem Domain	Engineers needed fast fixes to get planes flying again.	Lawmakers, clerks, auditors need fast policy analysis.
Core Data Sources	Service logs, repair manuals, parts DB, emails.	Legislative bills, statutes, public feedback, case law.
Integration Backbone	Endeca faceted search + metadata tagging.	Qdrant vector DB + Postgres RBAC.
Ontology Layer	Taxonomy + rules	Role-based filters + domain ontologies
Search & Reasoning	Ontology-driven contextual search	Retrieval-augmented generation + prompts
Inference Engine	Rule-based predictions	LLM reasoning
UI / Delivery	Engineers received “repair packs”: instructions, estimated repair time, expert contacts.	Lawmakers receive structured responses: bill context, fiscal impact, risks, prior examples.
Value Proposition	Reduce downtime, lower cost, faster customer support, consistent answers.	Faster policymaking, cost savings, improved legislative clarity, institutional memory.
Strategic Goal	Turn “exceptions” into “known cases”	Turn “policy questions” into “data-driven decisions”

🔑 TLDR

Client’s KL pattern = today’s Legislation → SCRY pattern.
Endeca faceted search has evolved into Qdrant embeddings.
Manual summarizers replaced by LLMs that self-summarize and reason.
The vision — turn chaos into predictability through knowledge — is identical.

2008 - Boeing Commendation

Revisiting my original blog and my Boeing project in 2008. I’d forgotten this letter... Technical Team Lead XXXXXXXX was one of my five managers. I accomplished quite a bit that year. We ripped off Seibel’s face, added a new one plus a queuing system in a “multi-tenant” design. Some excerpts

browardbroward horne

Multi-Tenant Systems - 1996 to 2018

My first application, Aspen [http://www.ugpti.org/research/projects.php?view=93&program=tssc], was multi-tenant. State agencies could alter forms, add fields, change printed reports and that’s why it succeeded. About half of my projects have been multi-tenant. -------------------------------------------------------------------------------- We used Adobe Flex and the Cairngorm framework for

browardbroward horne