Overview
This is a tool I built to fix real problems at my job as a Vendor Service Manager at Amerit Fleet Solutions. I track 40-100 fleet units across 6 food distribution centers, and the existing Smartsheet workflows were slow and manual. Nobody was going to fix that for me, so I started building something myself.
This tool has gone through three distinct phases — each one more capable, and each one more honest about who wrote what.
Phase 1 — Hand-Coded HTML Forms (Early 2024)
I started simple. Hand-coded HTML forms to speed up email generation and unit tracking. No frameworks, no APIs. Just me solving an immediate problem at my desk.
This was 100% my work.
Phase 2 — VSM-Tracker: Full-Stack MERN App (Oct–Nov 2025)
Built a proper full-stack application from the skills I was learning in school. 87 commits over 25 days, all mine, in the vendor-manager-web repo's predecessor (VSM-Tracker repo).
What I built myself:
- Node.js/Express backend with MongoDB
- JWT authentication with role-based access (admin/user)
- Email generation via Nodemailer with vendor contact management
- CSV import from Smartsheet and Maximo data sources
- Kanban-style dashboard for unit tracking
- Docker deployment on my Unraid homelab
- Dark/light theme system
- Analytics dashboard with charts
The commit history shows real development — setting up auth, debugging email integration, fixing duplicate unit merging, refining the UI over daily use. This was my code, written while I was taking CS courses and learning these technologies.
Phase 3 — AI-Expanded Dashboard (Jan–Mar 2026)
As I needed more advanced features, I started using Claude Code as a development partner. 56 commits in the vendor-manager-web repo.
What I directed and AI helped build:
- Smartsheet API integration with CORS proxy and caching
- Chart.js dashboards with 6 drill-down report types
- Claude AI integration for estimate document analysis (PDF/image extraction)
- 9 workflow input types with dynamic forms
- 3 report generators (weekly Teams summary, site email, customer update)
- Email generation with per-vendor contact picker
- User authentication backed by Smartsheet user store
- Vendor scorecard ranking system
- PDF analyzer tool and local file save utility
What's honest: I identified every problem, designed every solution, tested everything against real work data, and iterated daily. But AI wrote most of the advanced code. I directed the architecture, verified the output, and learned from the process.
The MERN Rebuild (Separate Project)
I also had AI scaffold a modern MERN stack rebuild (vendor-manager-mern repo, 4 commits). It's a reference architecture for eventual team-wide deployment. I'm not claiming I wrote that codebase.
Results
- Used daily at work for 2+ years
- Every vendor interaction at my job flows through this tool
- Reduced active out-of-service units from 150+ to manageable levels
- Automated reports that previously took hours
What I Learned
- How to identify real workflow problems and design solutions iteratively
- Full-stack MERN development (Phase 2 was my code)
- Working with APIs (Smartsheet, Claude) at increasing levels of complexity
- The difference between directing software development and writing every line yourself
- Docker deployment and homelab hosting for real daily-use tools
Next Steps
- Python/Flask rebuild of the dashboard without AI assistance, as a learning exercise
- Complete the MERN version's frontend to learn React patterns hands-on