Haven

Inspiration

Over 60 million people in the U.S. are neurodivergent, including individuals with ADHD, autism, anxiety, and more.

Most apps and systems today assume everyone experiences the world the same way ⚙️. But that’s not true.

For many people, simple things like walking to a coffee shop ☕ can be stressful because of:

Loud noises 🔊
Bright lighting 💡
Crowded spaces 🚶‍♂️

Around 85% of neurodivergent individuals deal with sensory overload daily, yet there are no tools that warn them ahead of time.

That’s why we built Haven 🚀.

Instead of just giving directions, Haven helps people move through the world in a way that feels comfortable, predictable, and safe 💙.

What it does

Haven is a navigation and daily support app built for people who are sensitive to their environment. It focuses on three main things:

Finding comfortable places 📍

Haven scores locations based on:

Noise 🔊
Lighting 💡
Crowd levels 🚶‍♂️

If a place looks overwhelming, it suggests calmer nearby options so users can make better decisions.

Helping with tasks 🗂️

Haven includes simple tools to make tasks easier, like:

Time blocking ⏱️
Checklists ✅
Body doubling (working alongside someone)

This helps users stay focused and reduce stress when completing daily tasks.

Real-time environment awareness 👓📱

Haven uses the camera and microphone to understand what’s happening around the user in real time:

Detects people to estimate crowd levels
Measures noise through the mic 🎤
Checks brightness using the camera

If things get too overwhelming, Haven speaks to the user 🗣️ and asks how they’re feeling. It can then suggest actions like taking a different route or pausing.

How we built it

Frontend: Built with React and Next.js for a smooth UI
Camera analysis: Uses TensorFlow.js to detect people and objects in real time
Audio: Measures noise levels using browser audio tools
Brightness: Estimates light levels from the camera feed

When something crosses a user’s comfort level:

AI generates a short check-in message
Text is converted to speech
User responds using voice
The system reacts accordingly

For navigation:

Google Maps is used for directions
Leaflet handles live tracking
GPS updates routes in real time

We also use a backend to handle heavier processing and improve recommendations.

Challenges we ran into

Performance Running camera analysis, audio processing, and UI updates at the same time slowed things down. We had to optimize carefully to keep it smooth.
Voice interaction Getting AI → speech → user response → action to work smoothly was tricky. If one step lagged, everything broke.
Maps integration We used multiple mapping systems, and they didn’t always work the same way. Making them feel seamless took extra effort.

What we’re proud of

Everything runs in the browser....no special hardware needed 💻

Real-time detection of crowd, noise, and lighting
A system that actually responds and talks to the user
Smart rerouting when environments become overwhelming

What we learned

Browsers today are way more powerful than expected
Combining multiple AI tools in real time is hard but possible
Accessibility is not just about visuals — it’s about how people feel and process the world

What’s next 🚀

We want to: Integrate with smart glasses (like Meta Ray-Bans) so the app works passively while walking Connect to wearables to track:
Heart rate ❤️
Body temperature 🌡️
Stress signals

This will help Haven understand not just the environment, but also how the user is feeling in real time, making support even more personal and accurate