projects · directed, not typed
The portfolio
One method throughout: I specify, test, and troubleshoot; the AI writes the code. The argument for why that works is in the case study. These are the other proofs.
RoseStorie
A multi-tenant family-dashboard SaaS, built and launched into production in spring 2026: tenant provisioning, Stripe billing, and a superadmin control plane, self-hosted behind Cloudflare. Grew out of a single-household Raspberry Pi kiosk dashboard.
Tournament platform rewrite
A ground-up rebuild of the tournament’s registration and web platform — systems I have run since around 2009 — with cutover planned for late 2026.
PlannerBuilder
My wife wanted a planner that did not exist commercially, so I built it for her. Then I automated its production: a custom-order, print-on-demand storefront with Stripe checkout and automated print fulfillment, deployed on Google Cloud Run. Feature-complete; launch pending.
Home lab & observability
The infrastructure everything above runs on, operated hands-on: pfSense and Proxmox on-premises with a self-hosted CI runner, a cloud droplet serving 49 domains over Tailscale, and a Grafana/VictoriaMetrics observability stack with custom exporters covering everything from DHCP drift to the house’s solar, battery, and smart-panel telemetry.
Purpose-built tools
A self-hosted sensor recorder, smart-electrical-panel analytics, a vehicle-telemetry bridge, an offline softball stat tracker, a nonprofit radio club’s website migration — small tools built because a purpose-built view beats a vendor app.
Status tags reflect honest state: “live” means running in production today; “pre-launch” means built but not yet selling or serving outside users.