Project Details

Betrak

Betrak takes a short assessment covering daily usage habits, sleep patterns and mental health, then runs the data through a machine learning backend to classify your addiction level with a confidence score. The results page breaks down your profile with interactive charts and gives personalized recommendations based on the prediction. Built with Next.js on the frontend and FastAPI + scikit-learn on the backend, with Google OAuth.

Features

Addiction level classification: Low, Medium or High with confidence score

Multi-step assessment covering usage, sleep and mental health

Interactive results with daily usage chart, intensity gauge and sleep correlation chart

Personalized suggestions based on prediction output

Assessment history stored per user in PostgreSQL

Tech Stack

Challenges Faced

First time using Python as a backend and training an actual ML model, bridging a FastAPI service with a Next.js frontend while handling data preprocessing, SMOTE balancing and model serialization was all new territory.

What I learned

How to connect a Next.js frontend with a Python/FastAPI backend across different runtimes and how a trained classifier actually turns input features into a real prediction with confidence.

Future Plans

Add a personalized progress tracker showing improvement or decline across sessions

Introduce a comparison view showing how a user's lifestyle stacks up against the healthiest profiles in the dataset.

Enable users to export their results and prediction history as a downloadable PDF report