Lepton Software · open-source sample2024 – 2025
Google RMI Demo
Production traffic intelligence platform for Google's Roads Management Insights program — dual-map comparison, time-replay, congestion analytics.
Module walkthroughs
Use case · Real-Time monitoring
Live anomaly detection across Paris ring road and Tokyo arterials — severity-ranked congestion alerts with one-click jump to the route on the map.
Use case · Historical analytics
Date range, day-of-week, and time-range filters with multi-city support — Boston, Gurgaon, and more. Average-delay and delayed-route counts aggregate live as filters change.
Use case · Route reliability
Pre-defined corridor analytics — Rome sightseeing, NYC airport corridors. Peak congestion, travel time, and 95% reliable time charted by hour. Quick Compare splits weekday vs weekend side-by-side.
Context
RMI is Google's Roads Management Insights program — agencies and consultancies use it to get high-resolution traffic and congestion data from Google's traffic graph. The demo I built is the official customer-facing showcase for the program, deployed by Lepton and open-sourced under the googlemaps-samples org.
It demonstrates the full RMI API surface (Roads, Geocoding, BigQuery Storage) wrapped in a production-grade React 19 + FastAPI app that customers can fork as a starting point.
What I built
- Interactive 3D maps with Google Maps Platform + Deck.gl: dual-map weekday/weekend and period-over-period comparison, time-replay autoplay, and color-coded congestion layers.
- Route reliability metrics — Planning Time Index, Travel Time Index, 95th-percentile travel times — computed with GeoPandas, Shapely, and Turf.js.
- Grid- and polygon-based urban congestion heatmaps with boundary analysis (postal codes, localities, admin areas).
- Frontend optimisations — lazy loading, code splitting, TanStack Query + Zustand — so interactions stay sub-second on city-sized datasets.
- Multi-stage Docker + Cloud Run with zero-downtime blue/green releases.
Impact
- Sub-second interactions even on city-scale datasets (~100k–1M road segments).
- Live and historical traffic insights composed from the Roads Management Insights API, Geocoding API, and BigQuery Storage API.