Working Systems · Live in Production

Not slides.
Working software.

These are live, operational systems — built, deployed, and operated by one engineer. Each demo opens in a dedicated environment. Interact with real data, real APIs, real results.

AI Governance Platform — Full Dashboard (v3)

Point it at any public GitHub repository and receive a scored governance compliance analysis in under 30 seconds — deterministic pattern analysis against real source code, no mock data, no AI-generated guesses. The scanner is one module inside the unified v3 governance cockpit: project onboarding, PMO dashboard, governance scorecard, gate manager, artifact generator, and a Framework Explorer covering NIST AI RMF, ISO 42001, EU AI Act, HIPAA, GDPR, and SOC 2. v3 adds a runbook library — DevSecOps, PM/Scrum, Workshop Guide, Policy as Code, and Software Factory — so every AI-generated commit is scanned against the same rule set as human-written code.

Live
Framework Coverage
NIST AI RMF · ISO 42001 · EU AI Act · HIPAA · GDPR · SOC 2 · ModelOps · AIOps · XDR · FinOps
Scanner
Up to 60 source files per scan · GitHub Trees API · subfolder targeting · per-pillar scoring · Critical / High / Medium / Info severity
Runbook Library (New in v3)
DevSecOps · PM/Scrum · Workshop Guide · Policy as Code · Software Factory
Program & Developer Tools
Project Onboarding · PMO Dashboard · Governance Scorecard · Gate Manager · Artifact Generator · Framework Explorer
Enterprise COTS App Rationalization

Multi-agent AI pipeline that scores an enterprise application portfolio against cloud migration criteria and produces auditable, per-app recommendations — dependency mapping, ROI math, ISV/EOL evidence, and a provenance table showing exactly which deterministic rule validated (or overrode) each AI recommendation. Paired with an interactive Digital Twin built on the same masked healthcare engagement data: a force-directed dependency graph exposing hidden cross-stack risk, and a 5-wave migration scenario simulator with TCO modeling, so stakeholders test migration waves before committing instead of discovering the risk in production. Every number traces back to a governance rule, not a model guess. Details masked for client confidentiality.

Live
Run Stats
36 clinical applications · 408 VMs · $24 total AI compute · 4.9 hrs runtime
Governance
22 deterministic validation rules · every recommendation provenance-tagged · zero hallucinated numeric outputs
Per-App Detail
5 tabs per app — Overview, Backlog, Provenance, Technical Details, Agents — with rationale, dependency graph, ROI breakdown, and a pass/fail provenance table
Digital Twin Scale
1,053 servers · 198 applications · 4 global data centers · real engagement data, masked for confidentiality
Dependency Risk
Force-directed graph exposes cross-stack dependencies invisible in spreadsheets — clickable blast-radius simulation
Migration Simulator
5-wave scenario planner with TCO modeling and infrastructure winddown — test assumptions before committing to a wave
Software Factory — AI Code, Governed by Construction

A premium add-on to the Governance Platform and MemoryOS: an AI code-generation pipeline — Ideation → Design → Development → Deployment → Operations — that writes application code against a build spec already encoding the applicable governance requirements, so the output is scanned-clean by construction, not by luck. Every AI-generated commit runs through the same scan engine as the Code Scanner above, and Refactor/Rebuild recommendations from the App Rationalization engine feed directly into this pipeline as its entry point. Consolidates two previously separate, working codebases — a 6R/7R rationalization engine and a discovery-to-audit build pipeline orchestrated with LangGraph. Governance Scanner and MemoryOS integration are still in progress, and no production deployment has shipped yet.

Live
5-Gate Pipeline
Ideation → Design → Development → Deployment → Operations — 2 human-in-the-loop approval gates minimum per build
Pipeline Stages
Discovery → assessment → requirements → code generation → testing → security scanning → quality gate → IaC → audit, orchestrated with LangGraph
Governance Tie-In
Generated code runs through the same scan engine as Code Scanner; Gate Manager blocks advancement on unresolved findings; required artifacts pre-populate from the build spec
Rationalization Tie-In
6R/7R portfolio scoring decides Retire / Retain / Rehost / Replatform / Refactor / Rebuild / Repurchase — Refactor/Rebuild recommendations are this pipeline's entry point
Sovereign Deployment
One model router targets public cloud AI, a client's own commercial cloud, or an AWS GovCloud boundary — decided once per component at Design time
Status
Consolidates 2 working codebases (rationalization engine + build pipeline) · 14 sample legacy apps for testing · Governance Scanner / MemoryOS integration in progress
MemoryOS — Enterprise Knowledge Ingestion Pipeline

A new enterprise data AI pipeline design: connects to just about any enterprise data source — chat, ticketing, wikis, document repositories, email, or a custom pipeline — through an extensible connector library, then runs every document through a 4-stage Bronze → Silver → Gold enhancement pipeline before it ever reaches a search index. Bedrock Guardrails and Claude are already the production AI layer today, called from Azure-hosted infrastructure — this page documents the full AWS-native port target, a design spec rather than a deployed system.

Live
Ingestion
Extensible connector library — pre-built adapters plus custom pipeline support — normalizes any source into a standard envelope schema
Stage 1 · Prune
Zero-inference, rule-based filtering discards system noise before any token-cost service runs — eliminates 70%+ of volume pre-AI
Stage 2 · Guard
Bedrock Guardrails screens denied-topic content; anything it intervenes on is archived with zero Claude tokens spent
Stage 3 · Classify & Enrich
Tool-forced Claude extraction (temp=0) adds confidence score, operational domain, and risk score to every surviving document
Stage 4 · Index & Serve
Embeddings feed a hybrid BM25 + vector index with Reciprocal Rank Fusion and semantic reranking, served tenant-scoped with PII masked pre-Bedrock
Historical-inator — AI Local History Platform

A location-aware generative AI product: drop a GPS pin anywhere and get an AI-narrated historical brief sourced from 60M+ digitized public-domain records (Library of Congress, National Archives, Smithsonian, DPLA) — tuned to 5 distinct audiences from casual visitors to K-12 educators and tourism boards. Built to qualify for NEH / IMLS cultural-heritage digital-access grant funding.

Live
Data Sources
60M+ records · Library of Congress · National Archives · Smithsonian · DPLA — all public domain, $0 access cost
Audiences
5 segments: consumers, K-12 educators, homeschoolers, tourism boards, cultural institutions
AI Pipeline
GPS capture (opt-in, read-only) → multi-source historical aggregation → generative narrative tailored to audience depth
Market
$9.8B US ed-tech market · $500M+ annual NEH/IMLS heritage grant funding
AI Pipeline Token Economics

A full token audit of a 7-agent enterprise AI rationalization pipeline — real token counts from AWS Bedrock Converse API usage fields, not estimates. Per-agent cost breakdown, retry economics, prompt anatomy, and a prioritized optimization roadmap across a 36-application healthcare portfolio run. Shows $23.97 baseline path to $7.63 with three targeted changes.

Live
Run Stats
3.05M tokens · $23.97 total · $0.67/app · 250 API calls · 4.9 hrs wall-clock
Agents
7 specialized agents · Telemetry · Dependency · Procurement · Provisioning · Synthesizer · Confidence Advisor · Portfolio Narrative
Analysis
Per-agent cost bars · token anatomy · retry rate by agent · prompt size vs. necessity · benchmark table
Optimization
7 prioritized optimizations · P1 code changes only · P1+P2 near-term: 68% cost reduction · 6 tokenomics principles
Enterprise App Design

How I approach enterprise application architecture, shown across the apps on this page — several of which are built, or actively being ported, across more than one cloud. Every component is a managed service chosen for compliance, zero operational overhead, and pay-per-use economics, with explicit cost and data-path rationale documented per cloud.

Live
Enterprise COTS App Rationalization · AWS
AI Governance Platform · AWS + Azure
MemoryOS Knowledge Pipeline · Azure + AWS
Historical-inator · Azure + AWS
Application Rationalization Digital Twin · AWS + Azure
Software Factory Pipeline · AWS + AWS GovCloud (in progress)
Park Whisperer Platform · Azure + GCP + AWS
Park Whisperer

A full-stack AI platform for theme park intelligence. Real-time crowd prediction, sellout event forecasting, satellite weather analysis, ML ride-wait models, and daily AI-generated social content — all running 24/7 on a multi-cloud stack across Azure, GCP, and AWS. Built and operated solo under $100/month.

Live
Infrastructure
Azure Functions · GCP Vertex AI · AWS S3 · Cosmos DB · Container Apps
AI Stack
Multi-agent RAG · GPT-4o · Bedrock Claude · GOES-16 satellite ingest · custom ML models
Output
Daily AI content to 3 social platforms · ride wait forecasts · live sellout alerts
Operating Cost
< $100 / month · 13 platform components · 25+ deployed pipelines
Available for New Engagements

Like what you just saw?

These are working systems, not slides — and I take on a limited number of freelance engagements through AI Pathfinder LLC to build the same kind for other teams. If one of these demos is close to your problem, let's talk about yours.

Email Me → More Ways to Connect →