Inspiration
Mental health challenges among students and young adults are increasing, yet most people don't recognize the warning signs until stress and anxiety have already become overwhelming. We wanted to create a tool that could help users become more aware of their mental well-being before reaching a crisis point. Inspired by the growing youth mental health crisis and the power of AI to identify patterns in data, we created Bloom—a platform that turns everyday habits and reflections into actionable mental health insights.
What it does
Bloom is an AI-powered mental wellness platform that helps users better understand their mental health.
When users first join, they provide information such as their average sleep time, study time, screen time, and stress level. Bloom uses a machine learning model trained on student mental health data to predict an anxiety score.
Users can then log daily journal entries through text or voice recordings. Voice recordings are automatically transcribed and combined with habit-tracking data and the predicted anxiety score. Using Gemini AI, Bloom generates a daily mental health score out of 10, along with personalized insights and recommendations.
The platform also includes:
- Daily mood and habit tracking
- Mental health trend visualizations
- Voice journaling with transcription
- Therapist recommendations based on location
- Stress-relief activities and games such as Bubble Wrap
- Personalized AI-generated wellness feedback ## How we built it We built Bloom using a modern full-stack architecture:
- Frontend: React Native, Expo, and TypeScript
- Backend: FastAPI and Supabase
- Machine Learning: Scikit-learn, Random Forest, XGBoost, and Linear Regression
- AI Analysis: Gemini API
- Database & Authentication: Supabase
- Location Services: Google Places API
- Voice Processing: Speech-to-text transcription pipeline
We trained a machine learning model using a student mental health dataset containing factors such as sleep duration, study hours, screen time, and stress levels. After preprocessing and evaluating multiple models, the best-performing model was deployed through a FastAPI endpoint that provides real-time anxiety predictions.
Gemini then combines these predictions with journal entries and daily habits to generate personalized wellness insights.
Challenges we ran into
One of our biggest challenges was transforming subjective mental health information into meaningful numerical insights. Mental health is highly personal and difficult to quantify, so balancing accuracy with responsible AI usage required significant thought.
Another challenge was integrating multiple systems—including machine learning, AI analysis, voice transcription, location services, and habit tracking—into a seamless user experience. Ensuring that all these components communicated efficiently while maintaining performance and usability was a major technical hurdle.
We also had to carefully design the application so that it supports users without presenting itself as a replacement for professional medical care.
Accomplishments that we're proud of
- Successfully training a machine learning model capable of predicting anxiety scores from lifestyle factors.
- Creating a beautiful, modern interface that makes mental health tracking feel approachable and engaging.
- Integrating AI-generated mental wellness analysis using real user data.
- Building voice journaling with automatic transcription.
- Providing personalized therapist recommendations based on location.
- Combining predictive analytics, journaling, habit tracking, and wellness tools into a single platform. ## What we learned Throughout this project, we learned how machine learning, artificial intelligence, and behavioral data can work together to create meaningful health applications.
We gained experience with:
- End-to-end machine learning pipelines
- AI prompt engineering
- Mobile application development
- Cloud databases and authentication
- Voice processing technologies
- Responsible AI design in health-related applications
Most importantly, we learned that technology can play a powerful role in helping people become more aware of their mental health when designed thoughtfully and ethically.
What's next for Bloom
Our vision is to transform Bloom into a comprehensive preventive mental health platform.
Future features include:
- Wearable integration (Apple Watch, Fitbit, Garmin)
- More advanced anxiety and burnout prediction models
- Personalized intervention plans
- School and university wellness dashboards
- Anonymous peer-support communities
- Therapist appointment booking
- Real-time stress detection from wearable data
- Multilingual support
- Predictive mental health forecasting
Ultimately, we hope Bloom can help people recognize mental health challenges earlier, build healthier habits, and access support before problems escalate.
Log in or sign up for Devpost to join the conversation.