Platform engineer · AI systems · Data analytics
Building tools that give developers (and game masters) superpowers.
stablepiggy.com · LinkedIn · GitHub
ZERO TRUST JIT Exec Grant Stablecoin Wallet Infrastructure & Tooling for Agentic Platform & Financial Agency: A multi-tenant developer platform where AI agents operate inside isolated, sandboxed environments — managing containers, vaults, knowledge bases, and deployment pipelines through a unified MCP protocol interface.
Napoleon, the platform's AI assistant, helps developers ship code through managed infrastructure — all within the org-scoped, security-audited system.
graph LR
subgraph Clients["Clients"]
Dashboard["Web Dashboard"]
Junie["JetBrains Junie"]
Codex["OpenAI Codex"]
Claude["Claude Code"]
Foundry["Foundry VTT Module"]
end
subgraph Platform["StablePiggy Platform"]
MCP["MCP Protocol Server<br/><em>150+ tools</em>"]
Vault["Encrypted Vault<br/><em>AES-256-GCM · scoped tiers</em>"]
Containers["Container Orchestration<br/><em>Docker · Caddy · per-org isolation</em>"]
Bundles["Composite Templates<br/><em>multi-container apps</em>"]
Napoleon["Napoleon AI<br/><em>multi-provider · tool loop</em>"]
end
Dashboard -->|REST + SSE| MCP
Junie -->|MCP| MCP
Codex -->|MCP| MCP
Claude -->|MCP| MCP
Foundry -->|WebSocket Relay| Napoleon
MCP --> Vault
MCP --> Containers
MCP --> Napoleon
Containers --> Bundles
class Dashboard,Junie,Codex,Claude,Foundry client;
class MCP core;
class Vault secure;
class Containers infra;
class Bundles app;
class Napoleon ai;
classDef client fill:#10161d,stroke:#22d3ee,color:#f8fafc,stroke-width:1px;
classDef core fill:#061826,stroke:#38bdf8,color:#f8fafc,stroke-width:2px;
classDef secure fill:#111827,stroke:#32d74b,color:#f8fafc,stroke-width:2px;
classDef infra fill:#071a12,stroke:#32d74b,color:#f8fafc,stroke-width:1px;
classDef app fill:#1f1a05,stroke:#facc15,color:#f8fafc,stroke-width:1px;
classDef ai fill:#0b1220,stroke:#22d3ee,color:#f8fafc,stroke-width:2px;
Built-in wallet system with a live testnet. Agents can hold balances, transfer funds, and settle transactions — real money movement in a sandboxed environment. Designed for testing payment flows, agent-to-agent commerce, and transaction-driven workflows before going live.
The OINKMOJI protocol and other platform documentation are published openly. Architecture specs, protocol definitions, and integration guides available at stablepiggy.com/docs.
2,800+ automated tests across backend and dashboard. CI runs on every pull request — typecheck, drift guards, migration safety, and security audit gates all enforced before merge.
16-phase security audit with zero critical findings. Full threat modeling, GDPR-compliant data lifecycle (erasure cascades, export endpoints), and incident response runbooks. The platform enforces encryption at rest, scoped access tiers, and container-level network isolation out of the box.
Composite template system lets you define multi-container applications with lifecycle-aware volumes — separating runtime services, persistent data, and independently-scaled sidecars. Deploy a full-stack app with one command.
Encrypted vault with personal, org, and platform tiers. Secrets are injected into containers at runtime, never exposed in chat history or logs. A built-in redaction engine scrubs credentials from stored conversations automatically.
Works with Claude Code, Codex, and JetBrains Junie out of the box. Nearly 200 custom MCP tools. The online chat interface supports Anthropic, OpenAI, OpenRouter, and Ollama — pick the model that fits your workflow.
Waitlist with seat-capacity management, honeypot defense, and tamper-resistant signup flow. Designed for a measured launch, not a free-for-all.
As a representation of a containerized integration, I build a module for FoundryVTT that lets StablePiggy connect to the game server and safely relay the data from the container to the LLM and back again (all org scoped). Napoleon can help game masters run tabletop RPG sessions in Foundry VTT with real module content (currently optimized for one of Foundry's most popular game systems, Pathfinder 2e). Any container run on the platform can be spun up and have integrations built to allow Napoleon to interact with it.
An end-to-end video production orchestration system built for a media production client. CVPO transforms story concepts into structured, edit-ready video projects — generating scripts, storyboard frames, voice-over takes, and b-roll selections, then spinning up cloud GPUs to do the inference with customized images of frontier models. Finally, packaging everything for import into Adobe Premiere Pro using a custom OTIO output format to sync images, video and audio pre-edit.
graph TD
Concept["Creative Concept<br/><em>idea · prompt · outline</em>"]
Script["Script / Story Generation<br/><em>narration · beats · scenes</em>"]
Breakdown["Scene & Shot Breakdown<br/><em>shot list · prompts · continuity notes</em>"]
Assets["Asset Planning<br/><em>characters · locations · props · style refs</em>"]
Images["Image Generation<br/><em>DALL-E · SD · HF models</em>"]
Frames["First / Last Frame Planning<br/><em>continuity anchors</em>"]
VideoGen["Image-to-Video Generation<br/><em>WAN 2 · Hugging Face models</em>"]
KenBurns["Ken Burns Quick Effects<br/><em>image → video clips</em>"]
Motion["Motion Graphics<br/><em>lower thirds · titles · overlays</em>"]
VO["Voice-Over Generation<br/><em>Gemini TTS · multiple takes</em>"]
BRoll["B-Roll Search<br/><em>Internet Archive · stock sources</em>"]
GPU["GPU Job Router<br/><em>RunPod · Google GPUs · local fallback</em>"]
RenderJobs["Render / Generation Jobs<br/><em>queued · tracked · reproducible</em>"]
Timeline["Timeline Assembly<br/><em>clips · audio · captions · overlays</em>"]
OTIO["OTIO Timeline Export<br/><em>OpenTimelineIO</em>"]
Package["Export Package<br/><em>media · metadata · prompts · timeline</em>"]
Premiere["Adobe Premiere Pro<br/><em>final edit / polish</em>"]
Concept --> Script
Script --> Breakdown
Breakdown --> Assets
Assets --> Images
Images --> Frames
Frames --> VideoGen
Images --> KenBurns
Breakdown --> Motion
Breakdown --> VO
Breakdown --> BRoll
VideoGen --> GPU
GPU --> RenderJobs
RenderJobs --> VideoGen
VideoGen --> Timeline
KenBurns --> Timeline
Motion --> Timeline
VO --> Timeline
BRoll --> Timeline
Timeline --> OTIO
Timeline --> Package
OTIO --> Premiere
Package --> Premiere
class Concept,Script,Breakdown concept;
class Assets,Images,Frames media;
class VideoGen,KenBurns,Motion generation;
class VO,BRoll support;
class GPU,RenderJobs compute;
class Timeline,OTIO,Package export;
class Premiere app;
classDef concept fill:#10161d,stroke:#22d3ee,color:#f8fafc,stroke-width:1px;
classDef media fill:#061826,stroke:#38bdf8,color:#f8fafc,stroke-width:1px;
classDef generation fill:#071a12,stroke:#32d74b,color:#f8fafc,stroke-width:2px;
classDef support fill:#0b1220,stroke:#22d3ee,color:#f8fafc,stroke-width:1px;
classDef compute fill:#111827,stroke:#facc15,color:#f8fafc,stroke-width:2px;
classDef export fill:#1f1a05,stroke:#facc15,color:#f8fafc,stroke-width:2px;
classDef app fill:#18181b,stroke:#94a3b8,color:#f8fafc,stroke-width:1px;
The system runs on Google Cloud Run with a React frontend on Cloudflare Pages, backed by Supabase (PostgreSQL). AI generation uses OpenAI for story and frame generation and Google Gemini for text-to-speech and vision analysis. Access is restricted to the client's team via Cloudflare Zero Trust with Google OAuth.
Reusable production blocks (camera setups, lighting rigs, character profiles, editing styles) let the team build a library of templates that carry across projects — consistent visual language without rebuilding from scratch every time.
An AI game master assistant for Foundry VTT — available on Docker Hub and pending Foundry marketplace review.
Napoleon reads real module content — actors, scenes, journals — from purchased adventure modules and answers questions, builds encounters, places walls and lighting, and manages session continuity. All from inside the Foundry interface, scoped per-user so your game data stays yours.
The module ships as a derived Docker image with version-tracked releases, per-org relay isolation, and clear error handling when something goes wrong (five distinct failure cases, each with an actionable notification instead of a silent failure).
A Chrome Extension that automatically replaces your real personal information with aliases when using AI chat services — ChatGPT, Claude, Gemini, Perplexity, and Copilot.
Your prompts go out with aliases, responses come back decoded. Profiles never leave your device. 750 passing tests. Free tier available, Pro at $4.99/mo. Its development shaped the zero-trust privacy approach behind StablePiggy's secret redaction system.
Founding member of a consultancy delivering data analytics and BI solutions.
$2.5M in inventory savings identified for a consumer goods client through predictive modeling and demand forecasting. Built automated reporting pipelines covering sales orders, production actuals vs. forecast, and material costs. Delivered executive dashboards and actionable recommendations that drove procurement and logistics decisions.
Currently preparing StablePiggy for launch by onboarding trusted users in batches. Also building a Java SaaS application with my data analytics team — JHipster foundation, transposed from a working TypeScript prototype into domain-driven Java architecture — that lets enterprise users safely query their own data in natural language. NLP to SQL with deterministic output.
What I actually use in production:
| Domain | Technologies |
|---|---|
| Platform & Infrastructure | Node.js, TypeScript, Express, SQLite (WAL), Docker, Caddy, pm2, GitHub Actions CI/CD |
| Enterprise & Backend | Java, Spring Boot (JHipster), domain-driven architecture, PostgreSQL, NLP-to-SQL pipelines |
| AI & LLM Integration | Anthropic Claude, OpenAI, OpenRouter, MCP Protocol, RAG with Voyage embeddings, multi-provider tool loops |
| Cloud & Deployment | Google Cloud Run, Cloudflare Pages, Cloudflare Zero Trust, Supabase (PostgreSQL), Vultr VPS |
| Security | AES-256-GCM encryption, Argon2id hashing, JWT with revocation, HMAC verification, CSP/HSTS, container egress firewalling |
| Frontend | React, Vite, Tailwind CSS, SSE streaming, WebSocket |
| Data & Analytics | Python, pandas, predictive modeling, BI reporting, demand forecasting |
| Protocols | MCP (Model Context Protocol), WebSocket relay, Stripe webhooks, GitHub App webhooks, Foundry VTT module API |
Data Analytics and Pipeline Architecture Design. Full stack development program at the University of Toronto. E-commerce experience (Shopify store management and Liquid code). Former Master Electrician — the kind of background where you learn that cutting corners on safety gets people hurt, which turns out to be a useful instinct when you're writing security audits and building vault encryption systems.
StablePiggy: stablepiggy.com ·
greg@stablepiggy.com
Other Inquiries: gregcbarker@gmail.com ·
LinkedIn
