I’m building SafeRoute AI, a mobile navigation platform that helps women and girls move through the city with greater confidence. The app combines an interactive map, a safety-scoring algorithm, and an AI assistant that responds exclusively to text commands. Drawing on live incident feeds and historical crime datasets, the backend assigns a numerical safety score to every street segment, then surfaces risks on the map as interactive data points the user can tap for detail. To complete the MVP, I need end-to-end development—from data ingestion right through to polished iOS and Android builds—plus a lightweight admin panel for managing incident sources and tuning the algorithm. Key deliverables • In-app AI assistant (text only) that accepts natural-language requests such as “Show me the safest route home.” • Dynamic routing engine that weighs distance, time, and safety to propose alternative paths. • Crime-visualization layer rendering interactive data points at multiple zoom levels. • One-tap SOS and real-time location sharing with pre-selected contacts (Twilio or similar). • Secure user accounts with encrypted location history and GDPR-compliant data handling. • Admin dashboard to monitor feeds, re-score areas, and push updates. Acceptance criteria • Safety scores recalc in under 3 s for a 5 km radius. • Map clusters behave correctly between zoom levels 10–17. • Live-share link refreshes every 5 s and terminates automatically at journey’s end. • 95 % unit-test coverage on core scoring logic. Preferred stack (flexible): React Native or Flutter, Mapbox SDK, Python/Node microservices, PostGIS, TensorFlow Lite for on-device predictions. If you’ve shipped location-based apps or worked with real-time data overlays, tell me how you’d tackle crime-data ingestion, route optimisation, and your timeline for an initial release.