FitLife

Nearly 697 million people worldwide are affected by diabetes, and heart disease remains the #1 cause of death globally. Yet clinical-grade, personalized health guidance is often expensive, inaccessible, or limited to hospital settings. Public health messaging is often generic and broad. We wanted to build something personal, data-driven, and actionable. FitLife was born from the idea that preventive healthcare should be proactive, intelligent, and accessible to everyone.

What it does:

FitLife predicts an individual’s cardiovascular disease risk and diabetes risk using personal health inputs

How we built it

Frontend: React + Vite Interactive UI for health metric entry and progress tracking Backend: FastAPI REST API endpoints connecting frontend to ML models ML: Trained on CDC NHANES population health data, Diabetes datasets for classification, Supervised learning models for risk prediction

Challenges we ran into

1. Data preprocessing: Real-world health datasets like NHANES are complex and messy. Cleaning missing values, normalizing features, and aligning variable formats required significant preprocessing.
2. Model Generalization: We had to ensure the models did not overfit specific subgroups and could generalize across populations.
3. Keeping the interface intuitive without oversimplifying health risk interpretation

Accomplishments that we're proud of

Successfully trained ML models grounded in real CDC population data Delivered instant, personalized risk assessments Built a clean full-stack architecture from frontend to ML backend Generated actionable health recommendations rather than just predictions Created a tool scalable to underserved and low-resource populations

What we learned

Real-world healthcare data is powerful but requires careful cleaning and interpretation. Predictive models are only valuable if users understand and trust them. Preventive health tools must combine accuracy, clarity, and empathy. Bridging clinical data and consumer usability is both a technical and design challenge.

What's next for FitLife

Expand the user base to improve model accuracy, robustness, and personalization Leverage longitudinal user data to continuously refine risk predictions Use logged nutrition and exercise data to dynamically adjust recommendations Introduce adaptive calorie and heart rate targets based on progress trends Build predictive “what-if” health trajectory simulations Continuously retrain models with anonymized real-world data for smarter insights

Built With

• claude
• cursor
• fastapi
• javascript
• python
• react
• vite