Jeremy Neely Text Jeremy

Web Apps & Dashboards · Production

Fleet Insights

Monthly repair reports that show each branch which trucks cost too much.

Built for: A regional truck dealer and service company that runs fleet maintenance for a commercial customer with dozens of branches

Repair invoices one month, every branch Peer match same type, year, make Claude review checks every report Branch reports one per branch + district Fleet Insights Screens are not shown: they contain a customer's real figures. Keep or replace verdicts for every unit, ranked against its peers.
A diagram, not a screenshot: the real reports contain a customer's figures.

The problem

Every branch of a large commercial fleet sends in repair and maintenance invoices each month. Branch managers had no simple way to tell whether their spend was normal for that equipment, which units drove the cost, or when a truck should be replaced instead of repaired. The old process ran on spreadsheets.

What I built

I built a four-stage Python pipeline that compares every unit with peers of the same type and age and writes one self-contained report per branch, per district and for management. Around it are a Report Builder desktop app for the program staff, a Claude review of every report before it goes out, and a guarded workflow that lets non-developers ask for report changes in plain English.

How it works

  1. Pull a month of repair and maintenance records through a secure MCP connection.
  2. Match each unit with similar units by year, make, model and type.
  3. Rank the units and branches that spend most above their peers, with keep or replace verdicts.
  4. Have Claude check every report against a written checklist.
  5. Publish one interactive HTML report per branch and district.
How Fleet Insights is wiredEach report is a self-contained HTML file: no login, and no live connection once it is open.
Comes inInside Fleet InsightsGoes out Repair & PM records a month, over secure MCP Vehicle roster year, make, model, type Peer matching same year, make, model Gap analysis spend vs peer rate Verdicts & signals keep or replace, rework Claude review checked against a list Branch report one HTML file, offline District report same, rolled up Gap ranking for management

Built with

  • Python
  • pandas
  • PySide6
  • Claude Agent SDK
  • MCP
  • HTML/CSS/JavaScript
Text Jeremy about this Next: SecureAIDB All 38 tools Texting opens your Messages app with a note already written. He replies himself.
Repair invoices one month, every branch Peer match same type, year, make Claude review checks every report Branch reports one per branch + district Fleet Insights Screens are not shown: they contain a customer's real figures. Keep or replace verdicts for every unit, ranked against its peers.
A diagram, not a screenshot: the real reports contain a customer's figures.

More of Jeremy's work

All 38 tools, by category