Featured · Independent project
I built a product to show how I work.
FleetPulse is an independent prototype: a fleet management platform, twelve connected modules, simulated data. It sits where four parts of my background meet. Years in the domain told me what to build. Product thinking shaped the workflows. UX decided what each screen had to make obvious. AI got me from thinking to something interactive in days instead of a quarter.
FleetPulse
847 vehicles, 267 drivers, 6 regions, 14 tax jurisdictions. Live telemetry, hours-of-service compliance, maintenance and work orders, driver inspections, licensing, routing, cost per mile, and quarterly fuel tax. Simulated data, no backend. Go click around.
Product decisions
- A fleet manager's job is a queue, not a dashboard. Everything in this category ships an alert list that tells you what happened. The Action Center tells you what to do, with the action attached: pull this unit from service, ground this driver, call these two before dispatch.
- One record, five places. A driver inspection report shows up on the vehicle, on the driver, in the maintenance queue, and against the duty log, because those are the four places somebody needs it. An open out-of-service defect disables the Dispatch button on the vehicle record.
- Two sources that should agree. Fuel tax is filed on mileage, and mileage comes from both telematics and the odometer. The tax module reconciles them per vehicle and flags anything over two percent before the return goes out. The inspection log and the duty log get the same treatment, since a pre-trip inspection should leave a mark in both.
- Approval authority is separate from edit access. A maintenance manager can open a work order but cannot approve a repair over $2,500. Spending limits are a financial control, so they are modeled as their own permission rather than a step in a workflow somebody can route around.
UX decisions
- The first thing you see is a decision. The home screen opens with the three things that need you right now and lets you clear them. Charts sit underneath. Ninety seconds before the yard opens, you still leave knowing the right things.
- The navigation carries the workload. Live counts sit on the sidebar itself, so four hours-of-service violations and eighteen overdue services are visible before you click anything.
- Errors say what still works. When the billing integration fails, the panel leads with what broke, what is unaffected, and how long before it costs money. That is the question the person reading it actually has.
It is a prototype with simulated data, not production software. What I wanted to show is how quickly I can get from knowing a domain to something you can click and argue with.
Selected screens
Tap any screen to open it full size.
