✨ Overview
Aurora is a mobile app for mental well-being tracking, built during graduate research at NECSTLab (Politecnico di Milano). Users share short voice reflections daily; the app tracks them across eight well-being dimensions and uses an LLM to suggest activities. The system’s interaction model and response quality were validated through a formal 27-participant user study conducted over multiple weeks.
🏗️ Technical Architecture
Mobile Layer
- Built with React Native and Expo for cross-platform deployment.
- Interaction flow centered on daily voice-based sharing with structured follow-up.
Backend
- Serverless infrastructure using AWS Lambda functions.
- LLM integration via AWS Bedrock with a fine-tuned Meta-Llama-3.1-70B instance.
Data and Privacy
- Structured data collection across eight well-being parameters.
- Historical record maintained to support focused follow-up with mental health professionals.
- Data sharing with professionals requires explicit user consent.
👥 User Study
Collected and analyzed data from 27 participants across multiple weeks, covering interaction patterns and LLM response quality.
🏆 Outcomes
Won Hack the NECSTCamp, a competitive 36-hour AI mobile application challenge, delivering a complete end-to-end product: formal user study, serverless LLM backend on AWS, and a cross-platform mobile app ready for real users.