SheSteps

• 

  maps

• 

  home

• 

  call

Inspiration

Many women feel unsafe when walking or running alone, especially in unfamiliar or poorly lit areas. We wanted to build a proactive tool that not only helps users stay aware of their surroundings, but also deters potential threats before they happen. We have personally experienced this, and tried to include features that we thought would be the most helpful to us.

What it does

SheSteps combines a machine learning risk score with a conversational AI companion. The mobile app analyzes real-time location features to predict the safety of the current route, while the voice agent holds a natural conversation—making it clear the user is not alone and can call for help if needed.

How we built it

We engineered a Gradient Boosting Classifier trained on geographic and environmental features to assign a safety risk score. This integrates with a React Native mobile interface that displays risk alerts. We used speech recognition and text-to-speech APIs to power the AI voice companion, coordinating active voice engagement when risk is elevated.

Challenges we ran into

Finding the right features to quantify “safety” was difficult due to subjective and location-specific variables. Balancing responsiveness with user privacy and ensuring smooth real-time voice interaction also required multiple iterations.

Accomplishments that we're proud of

We successfully built something fuses risk prediction and an AI-driven safety presence—two technologies rarely combined in this context. Our system adapts conversation behavior dynamically as risk changes.

What we learned

We deepened our skills in geospatial feature engineering, ML evaluation, and real-time mobile system design. We also learned how important user psychology is in designing tools for safety and comfort.

What's next for SheSteps

We plan to incorporate crowdsourced safety reports, SMS emergency contact notification, and other emergency escalation features. We also aim to expand our model with larger datasets and personalization, helping more people feel confident moving through the world.

Built With

• claude
• google-directions
• google-maps
• javascript
• livekit
• python
• react
• scikitlearn
• typescript