Midnight Mile

"You're never truly walking alone."

Inspiration

Walking home at night should feel safe but for many women it does not. After speaking with friends and reflecting on personal experiences, we noticed a gap: current map apps help with navigation but not safety. We imagined a solution where safety is not a feature but the foundation.
Thus, Midnight Mile was born: a map-based app that offers AI-powered safe walking at night, blending real-time data, trusted help, and immersive reassurance.

What it does

Midnight Mile provides a map-centric app experience where users enter a destination and receive the safest walking route based on crime data, lighting, and foot traffic. It highlights nearby safe spots such as police stations, hospitals, and 24/7 stores. Users can activate an AI voice companion that simulates a trusted friend walking alongside them, periodically checking in, responding to distress cues and listens to you for rerouting to a safe spot. Upon arrival, users confirm safety, or alerts are sent automatically to trusted contacts.

How we built it

Core Stack

Google Maps APIs:
- Directions API for road-to-road route rendering
- Places API to fetch safe zones (police stations, hospitals, 24/7 stores)
- Maps JavaScript API for Embedding custom map to our website
ElevenLabs API:
- Used for real-time voice streaming
- Custom voices that simulate a calm male companion
- Background noise analysis for distress keyword detection (planned)
Front End:
- Next.js
Database:
- Supabase (PostgreSQL) -LLM:
- Gemini
Voice and Safety Logic:
- Voice asks “Are you okay?” periodically
- Sends alerts to trusted contacts if no response or panic keywords detected
- “Safe Now” auto-check-in on arrival using geofencing

Challenges we ran into

Streaming voice in real time without latency on no-code platforms
Making the AI companion feel natural and human, not robotic
Obtaining reliable, up-to-date safety data for cities like Delhi
Balancing immersive AI features with a simple, clean UX

Accomplishments that we're proud of

Successfully integrated multiple Google Maps APIs for safety-focused navigation
Developed a lifelike AI voice companion using ElevenLabs that enhances user comfort
Designed an intuitive, map-first UI centered on real-time safety information
Created a scalable user flow combining destination input, route guidance, and safety check-ins

What we learned

The importance of combining real-time data with empathetic AI to improve safety perception
Challenges of working with voice streaming APIs on no-code platforms and ways to overcome latency
How to design for emotional impact using minimalistic but meaningful UI elements
Practical insights into integrating AI and geospatial data for social impact

What's next for Midnight Mile

Implement panic trigger detection using real-time audio analysis
Expand safety data coverage to more cities worldwide
Personalize the AI companion’s language, voice, and emotional tone for diverse users
Enhance immersive features such as Street View integration for richer navigation experience
Have verified safe spots other than public such as people who can register to provide help