Thrive
A mobile app for young adults that estimates depression levels from a voice recording.
- Client
- Bangkit Academy capstone
- Year
- 2024
- Role
- Project lead, backend, ML deployment
- Stack
- Hapi.js, Flask, Librosa, Cloud Run, Firestore
The problem
Thrive set out to help young adults notice depression early, from something as simple as their voice.
What I built
I led the scope, budget and timeline, and the backend: a Hapi.js REST API with authentication, sessions and audio processing, and a Flask service that uses Librosa features to run the model. Everything deploys on Google Cloud through Cloud Build.
The outcome
A working mobile app that turns a voice recording into an estimated depression level, built as the capstone of Bangkit Academy, a program that selected its cohort from 57,000+ applicants.
How it works
- 01Recordingthe user speaks
- 02Hapi.js APIauth, sessions, audio
- 03Flask + Librosafeatures and the model
- 04Estimatestored in Firestore
Details
- Led scope, budget and timeline for the team.
- A Hapi.js REST API with authentication, sessions and audio processing.
- Models served on Google Cloud with CI/CD from GitHub through Cloud Build.
Results
- Voicein, depression estimate out
- CI/CDCloud Build from GitHub
