Inspiration
Student burnout is rising incredibly fast! We wanted to build a tool that doesn’t just measure stress but helps students understand why they may be burning out and what they can do about it. Our goal was to turn raw wellbeing data into clear, actionable insights that actually help people.
What it does
MindGauge predicts a student's burnout risk by analyzing their academic load, mental health indicators, lifestyle habits, and professional wellbeing scales. It outputs an evidence backed burnout score, highlights any concering risk factors, and uses Gemini AI to generate personalized recommendations that explain why the suggestion matters for that specific student.
How we built it
We started by analyzing the CARRAD dataset containing 1790 students' mental health and burnout metrics. Through data analysis, we identified 12 key predictors of student burnout, including depression scores (CESD), anxiety levels (STAI-T), study hours, sleep quality, and social support networks.
Built a Random Forest Regressor to predict all three dimensions of the Maslach Burnout Inventory (MBI-SS): Exhaustion, Cynicism, and Professional Efficacy. Performance: Achieved strong predictive accuracy with Depression (CESD) emerging as the strongest predictor (46.3% feature importance) Validation: Implemented clinical threshold-based categorization (Low/Moderate/High risk) aligned with MBI-SS research standards
We iterated through multiple model versions, starting with a simpler 5 feature model and expanding to 12 features based on burnout literature. We also tested recommendation formats, ultimately settling on numbered lists with a "WHY:" statement/explanation that performed best in comprehension tests.
Challenges we ran into
- Restructuring a complex set of psychology data with multiple scoring metrics
- Modeling three separate burnout dimensions instead of a single score.
- Getting Gemini AI to produce consistent, personalized recommendations that directly reference user data.
- Keeping the UX simple despite handling many wellbeing variables.
Accomplishments that we're proud of
- Successfully predicting all three MBI burnout dimensions with strong performance.
- Building meaningful, personalized recommendations backed by actual psychology research.
- Designing a clean dashboard making difficult mental health data understandable.
- Creating an app that feels genuinely helpful and not just a ML demo.
What we learned
We learned how complex burnout really is and why most tools often fail. We also improved our skills in dataset preprocessing, model tuning, prompt engineering, and designing mental health based oriented UX that’s empathetic and responsible.
What's next for MindGauge
- Adding ongoing tracking so students can monitor burnout over time.
- Expanding to include passive data
- Deploying the tool for universities to help students in real time
- Using LLMs to provide long-form guidance and evidence-based mental-health strategies.
Log in or sign up for Devpost to join the conversation.