GlucoGuard

Inspiration

We were inspired by a simple problem: knowing your glucose level does not always explain why it changed. The same meal can affect different people differently, while sleep, physical activity, meal timing, and previous glucose patterns can also influence glucose response. We wanted to build a solution that helps people understand their personal glucose patterns, rather than simply recording numbers.

What it does

GlucoGuard is an AI-powered web application that learns an individual's glucose-response patterns and predicts the risk of a post-meal glucose spike.

Users can record their meals, glucose readings, meal timing, sleep, and physical activity. GlucoGuard analyzes the current information along with historical patterns and provides a Low, Medium, or High Spike Risk.

Its key feature is the Personal Spike Fingerprint, which learns how a particular user tends to respond to different meals and lifestyle patterns. The system can also suggest small, practical meal adjustments that may help reduce predicted spike risk.

How we built it

We built GlucoGuard as a web-based application using:

HTML, CSS, and JavaScript for the frontend
Python and Flask for the backend
Scikit-learn for machine learning
SQLite for data storage
JavaScript charts for visualizing glucose patterns

The overall workflow is:

User Data → Pattern Analysis → ML Prediction → Spike Risk → Personalized Insight

We designed the system to focus on Indian and regional meals, with the goal of making personalized glucose awareness more relevant to local users.

Challenges we ran into

One of our main challenges was avoiding overly general predictions. A meal that produces a high glucose response for one person may not have the same effect on another.

We therefore focused on learning from individual historical patterns instead of treating every user the same. Another challenge was presenting AI predictions in a simple and understandable way while ensuring that the application does not present itself as a medical diagnosis or replace professional medical advice.

Accomplishments that we're proud of

We are proud of building a complete concept that combines machine learning, personalization, healthcare, and a usable web interface.

We especially focused on the idea of a Personal Spike Fingerprint, allowing the system to learn from an individual's previous responses rather than relying only on generic food information.

We are also proud that the project can be developed as a web application without requiring custom hardware, making the initial prototype more accessible and feasible.

What we learned

We learned that building healthcare AI is not only about prediction accuracy. Personalization, explainability, usability, and responsible communication are equally important.

We also learned how machine learning can combine multiple factors—such as meals, glucose history, sleep, activity, and timing—to identify meaningful individual patterns.

What's next for GlucoGuard

Our next step is to improve personalization as more user data becomes available.

Future versions could include:

CGM integration for automatic glucose readings
Wearable integration for sleep and activity data
A larger Indian and regional food database
Improved meal-adjustment recommendations
Long-term personal glucose trend analysis
Optional caregiver and healthcare-professional reports

Our long-term goal is to move glucose awareness from simply asking "What is my sugar level?" to understanding "What patterns affect my glucose, and what small changes could help?"