SwaraIbu: AI-Powered Voice Based Mental Health Support for Mothers
An AI-powered mobile application that analyzes voice input to detect emotional distress in postpartum mothers and provide early mental health support, awarded 1st place at the SLASHCOM Hackathon.
Understanding the Context
Overview
The Problem
Postpartum depression often goes undetected due to limited emotional support, low awareness of early symptoms, and the absence of accessible mental health tools for mothers.
The Purpose
SwaraIbu aims to provide early emotional support by analyzing voice input to detect stress and critical conditions, enabling timely intervention and companion involvement.
My Contributions
Responsibilities
Developed the AI crisis check API for detecting emotional distress from voice input
Integrated AI analysis results into the mobile application workflow
Implemented Gemini AI integration on the mobile app for personalized responses
Collaborated with backend and mobile developers during a fully remote hackathon
Technologies Used
Tech Stack
📱 Frontend
⚙️ Backend
🤖 AI/ML
🛠️ Tools
Core Functionality
Features
Voice Recognition & Analysis
Analyzes voice input to detect emotional stress and potential mental health risks in postpartum mothers.
Emergency Alert
Triggers alerts and critical warnings when high risk emotional states are detected.
Mood Tracking
Tracks emotional changes over time to help users and companions understand mental health patterns.
Companion Access Mode
Allows trusted companions or family members to view summarized emotional conditions.
Visual Design
UI Showcase
Interface design and user experience highlights
Mother’s dashboard showing AI voice recording, analysis results, and emotional history.
Technical Journey
Challenges
Developing and deploying a machine learning API for the first time within a tight hackathon timeline.
Learned the fundamentals of deploying AI-based APIs and handling model inference through backend services.
Designing a crisis detection mechanism using pre trained NLP models without fine tuning, under a tight hackathon timeline.
Learned how to adapt pre trained models with domain specific datasets, semantic similarity thresholds, and text normalization to detect crisis conditions effectively.
Coordinating development during a fully remote hackathon while simultaneously preparing for KKN deployment.
Improved remote collaboration, time management, and task prioritization under pressure.
Results & Achievements
Impact
Measurable outcomes and project success metrics
Awarded 1st Place at SLASHCOM Hackathon
Delivered a functional voice based AI mental health prototype
Successfully integrated AI APIs with a mobile application
Next Steps
Future Plans
Improve voice emotion detection accuracy through expanded and more diverse audio datasets
Conduct usability testing with postpartum mothers and mental health professionals
Enhance AI deployment stability and scalability for real-world usage
Integrate wearable or sensor-based data for more comprehensive stress detection
Add multilingual support to reach a wider range of users