Problem
Dealer quotes depended on spreadsheets, PDFs, and a lot of manual work. It worked, but it wasn't consistent, and keeping pricing current across hundreds of dealers was hard.
What I built
I designed, built, and operate Start Blue. It holds the full tractor inventory and configurations, so anyone can configure a tractor and get a quote in a few clicks. Dealerships sign in with their own accounts to save quotes and apply their own pricing. More than 3,000 quotes have been generated.
Architecture
- React frontend, Django backend, and PostgreSQL, with object storage for files
- Three-node Kubernetes cluster on Linode/Akamai
- GitOps delivery: GitHub Actions builds images, Flux CD reconciles the cluster from Git
- SOPS for encrypted secrets in the repository
- Traefik and cert-manager for ingress and automated TLS certificates
- Automated restic backups of application data and cluster configuration
Operational concerns
Writing the application was only part of the job. I also own the container images, deployments, backups, PostgreSQL, monitoring, DNS, and TLS. Running a database inside Kubernetes means planning for persistent data, backups, and schema changes, not just restarting containers.
Impact
Start Blue has been in production for over two years without a major incident. I presented it at LS Mtron's headquarters in South Korea, including to the Global CEO, and the work was recognized with the 2025 CEO Award for Employee of the Year. I'm now working with engineering teams there to adapt it for multiple languages and other global markets.