How I built a production-ready pipeline that creates accurate digital basemaps of urban road networks — the foundation for measuring real travel times across Israel's major cities.
Developed as part of a flagship strategic-planning project for the Israeli Ministry of Transport, in collaboration with Israel's leading transport agencies.
“Every travel-time measurement is only as good as the map underneath it. This pipeline makes sure that map is right.”
To measure how traffic really flows through a city, you first need a precise digital map of every road segment — in each direction of travel. Off-the-shelf maps aren't accurate enough for that. So I built a system that creates these 'basemaps' from scratch, for Tel Aviv, Jerusalem, Haifa and Be'er Sheva.
The pipeline loads the city's road network and breaks it down into directed segments — a two-way street becomes two separate one-way links, each with its own unique ID.
Each segment is then compared against real-world driving routes. The system automatically measures how well the two geometries match, and flags every mismatch.
A dedicated visual review interface lets an analyst inspect each flagged segment on an interactive map, fix it with a few clicks — adjust endpoints, add waypoints, re-check — and approve it. Keyboard shortcuts make reviewing hundreds of segments fast.
The result: a validated, export-ready basemap that feeds the national travel-time measurement system.
A walkthrough of the basemap creation and review workflow.
I built this entire pipeline — the data processing, the interactive review interface, the automated test suite, and the production deployment — using AI-native development workflows with tools like Claude and Codex. AI didn't just write code snippets; it powered the full cycle: architecture, implementation, automated testing, and production hardening. This is exactly the workflow I teach in my AI lectures and consulting.
Transport agencies make multi-million shekel decisions based on travel-time data. An inaccurate basemap silently corrupts every measurement built on top of it. This pipeline turns a months-long manual mapping effort into a fast, repeatable, quality-controlled process — already in production use across four metropolitan areas.
Want AI-native workflows like this in your organization?