HelpHelp

Inspiration

Walking home alone at night. Waiting for a ride in an unfamiliar area. Feeling uneasy in a rideshare. Millions experience these vulnerable moments daily, with women, night-shift workers, and students at the highest risk. Research shows that attackers are significantly less likely to engage with someone who appears to be on a call, as it increases the risk of exposure. Inspired by this data and the need for hands-free emergency solutions, we created HelpHelp—a smart AI-powered safety app that proactively deters threats, records real-time evidence, and alerts authorities when danger arises.

What it does

AI-Powered Fake Call – The app starts a realistic AI-generated conversation when triggered, creating the illusion that the user is on a live call.

Wake Word Detection – Saying "help help" activates emergency protocols, discreetly recording audio, tracking location, and notifying authorities.

Automatic Emergency Response – Instantly sends the user’s location and recorded evidence to emergency contacts and law enforcement.

Smart Integrations – Uses email, SMS, and push notifications to alert authorities and contacts, ensuring swift assistance.

How we built it

We developed HelpHelp using React (Typescript) and Expo for seamless cross-platform functionality, integrating cutting-edge AI with Gemini for natural-sounding fake calls. EmailJS handles emergency contact notifications, while Firebase stores location and voice data securely. Wake word detection was trained using TensorFlow.js, ensuring fast and accurate voice recognition even in noisy environments as well as browser enabled text to speech.

Challenges we ran into

Wake Word Accuracy – Optimizing the AI to detect "help help" with high precision while avoiding false triggers.

AI Voice Realism – Making the AI-generated fake call sound convincing to deter threats effectively.

Fast & Secure Data Handling – Ensuring real-time transmission of location and voice data without latency or privacy risks.

Cross-Platform Performance – Maintaining smooth functionality across both iOS and Android, despite different hardware constraints.

Accomplishments that we're proud of

Successfully implementing real-time wake word detection with low latency.

Making something reliable, and creative in 8 hours.

Creating a lifelike AI-generated call experience that deters potential threats.

Seamlessly integrating location tracking and emergency notifications for rapid response.

Designing an intuitive and accessible hands-free safety solution that empowers vulnerable communities.

What we learned

The importance of human psychology in safety apps—attackers avoid people who appear engaged.

How to optimize AI wake word detection for speed and accuracy.

Best practices in secure data transmission and privacy protection for real-time emergency reporting.

The value of user experience testing in making the app intuitive and effective for people in distress.

What's next for HelpHelp

Video Call Feature – Users can opt to start a live video call with emergency contacts or authorities for better situational awareness.

High-Risk Area Mapping – Aggregating anonymous location data to identify and highlight high-risk zones, helping users navigate safer routes.

Smart Contact Escalation – If the first emergency contact doesn’t respond, the app will automatically notify the next in line.

Wearable Integration – Compatibility with smartwatches for even faster activation.

More Wake Words – Customizable phrases for personalized emergency triggers.

AI Sentiment Analysis – Detecting stress or fear in a user’s voice to preemptively activate safety features.

Built With

• css3
• emailjs
• gemini
• react
• typescript