Auto-scheduler
In production · Sole engineer
A custom engine that plans a full day of pre-planned medical rides in about three minutes.
I built it alone: a custom engine in C#/.NET. It works in passes: seeding, construction, then improvement, and every pass uses one shared feasibility checker. It places wheelchair and ambulatory riders on a mixed fleet, inside time windows and hours-of-service rules, and carpools riders where it can. One solve runs millions of calculations. The company's schedulers still review the plan.
About 10 months to build · in production since mid-2026
- C#
- .NET
- Scheduling and routing
- Feasibility checking
A simplified illustration of the scheduling problem, not the real engine. Synthetic data: a sample city and made-up riders.
- ~1,900rides a day, across about 250 vehicles
- ~3 minper solve
- ~$168Ka year saved in scheduling labor, about 6 full-time schedulers' hoursEstimated by Ryan
- ~$225Ka year in empty-mile savingsEstimated by Ryan, from planned figures at owner prices
| Measure | Scheduler | Team's plans |
|---|---|---|
| Empty miles per trip | 10.9 | 12.0 |
| On-clock minutes per trip | 75.3 | 78.4 |
| In-scope trips placed | 99% |
Lower is better for miles and minutes. The test compares plans, not measured on-time performance.




