2025–2026
Automated verification for a 30-day revenue service demonstration
A daily performance picture produced overnight, replacing roughly eighteen person-hours of manual analysis with a forty-minute automated run.
- Python
- TypeScript
- ATS data
- Data warehouse
- Scheduled jobs
The problem
Revenue Service Demonstration is the regime that determines whether a line is fit to carry passengers. Producing the daily performance picture by hand consumed roughly eighteen person-hours across three people, and it arrived after the morning review — so the meeting was spent reconstructing what had happened rather than deciding what to do about it.
What I did
I built a web application over Automatic Train Supervision and data-warehouse feeds, in Python and TypeScript with AI-assisted development, running on an automated overnight schedule. The design constraint was that the analysis had to be complete and trustworthy before the morning review, without anyone having to be awake for it.
What changed
The full daily picture landed before the morning meeting every day. Roughly eighteen person-hours of manual analysis became a forty-minute automated run, and the demonstration was managed against decisions rather than reconstructed from logs.
What I took from it
The value was never the time saved. It was that the conversation changed — from what happened yesterday to what we do about it today. That is usually the real return on automating an analysis.