Personalized Mental Health Campanion Application - MindMate
Feynn Labs
05 months
Machine Learning
My Approach:
MindMate was a proposed AI-powered mental health companion app concept. My role was research and technical proposal: I authored a technical paper and product report exploring how a hybrid recommendation system could power personalized mood tracking and content recommendations for a mental health app, and evaluated candidate recommendation approaches. The mobile application itself was built by a separate development team; my contribution was the research, system design proposal, and recommendation model evaluation, not the app implementation.
What I Researched
I explored how personalization could work in a mental health context, focusing on three recommendation approaches:
Content-based filtering — matching users to resources based on their own activity and mood history
Collaborative filtering (SVD-based matrix factorization) — matching users to resources favored by similar users
A hybrid model combining both approaches
Evaluation
Due to the sensitive nature of mental health data, I evaluated these approaches using simulated user-content interaction data rather than real user data. I compared the three approaches using Precision@5, Recall@5, and RMSE:
Model | Precision@5 | Recall@5 | RMSE |
|---|---|---|---|
Content-Based Filtering | 0.61 | 0.58 | 0.92 |
Collaborative Filtering (SVD) | 0.67 | 0.63 | 0.85 |
Hybrid Model | 0.74 | 0.71 | 0.78 |
The hybrid approach outperformed both standalone methods across all three metrics, supporting the case for a hybrid recommendation architecture in the product proposal.
Product and Market Research
Beyond the technical evaluation, I authored a product report covering:
Market analysis and customer segmentation for mental health apps
Core feature specification (mood tracking, personalized recommendations, community support, goal tracking)
Privacy and regulatory considerations (GDPR, CCPA, HIPAA)
Monetization strategy options (freemium, in-app purchases, partnerships)
Outcome
This research informed the technical direction for the MindMate concept, specifically the case for a hybrid recommendation approach over a single-method system. The actual mobile application was developed separately by Feynn Labs' engineering team.
