15+ apps, one developer, one year

The problem

When I joined Green Bay Converting, software was something the company bought. A year later it runs on an internal platform of 15+ production applications covering finance, HR, operations, and manufacturing — designed, built, and deployed largely by one developer, on infrastructure the company owns — eliminating $100K+ a year in software licensing.

What I built

• The infrastructure. A production Linux/Docker environment, provisioned and secured from scratch. Self-hosted and containerized, with no dependence on external cloud subscriptions — the platform's entire running cost is the hardware it sits on.
• The architecture. Every application authenticates through a centralized, MSAL-backed API gateway; database access is least-privilege throughout, so no app talks to SQL directly. Full CI/CD in Azure DevOps means every application builds, tests, and deploys automatically on every change.
• The AI layer. A custom MCP (Model Context Protocol) server connects AI tooling directly to internal apps and APIs. A retrieval-augmented generation assistant — Azure AI Search and the Claude API over a custom document-ingestion pipeline — gives machine operators instant, sourced answers from equipment manuals, on the floor, without hunting through binders.
• The handoff. A second developer has been trained on the full platform and on the AI-assisted workflow that built it. The platform doesn't depend on me — the model is repeatable.