Inspiration
Caregivers for Alzheimer’s patients often face the overwhelming task of tracking subtle, daily fluctuations in behavior, nutrition, and incidents without a structured way to share that data with doctors. This app was created to bridge this gap between home-based care and clinical oversight, creating a tool that transforms the exhausting process of manual observation and monitoring into actionable, data-driven insights that improve the quality of life for both patients and their families.
What it does
AlzheimerFairy Care System is a centralized HealthTech platform that allows caregivers to record daily "Check-ins" covering mood, cognition, diet, and medical incidents. The system uses a visual Care Calendar with color-coded alerts to highlight health trends and features a predictive engine that analyzes multivariate health markers to project symptom severity. Finally, it generates high-fidelity, printable 30-day clinical reports that caregivers can hand directly to physicians for more informed medical decision-making.
How we built it
The application is built on a modern Next.js and Typescript frontend, styled with Tailwind CSS and Lucide React for a mobile-first, responsive user experience. Supabase was used for secure data storage and user authentication, while the analytical backbone is powered by a Python (Flask) backend hosted on Render. Our predictive modeling utilizes pandas, NumPy. and Scikit-learn (specifically Random Forest algorithms with custom feature engineering) to process data points across physical and behavioral compliance markers.
Challenges we ran into
One of the primary challenges was designing a system that felt lightweight while remaining robust enough for clinical use, especially on a short deadline. Specifically, I had to make sure the Random Forest model could give meaningful projections based on a set of subjective "Check-in" data. Setting up each page to help caregivers in an intuitive and user-friendly manner, while offering detailed features and tasks, was a tricky aspect to balance. Additionally, managing the hosting and deployment using Vercel and the Render (through Gunicorn, Flask API) also required careful configuration to ensure low-latency responses for the ML-predictive trend analysis.
Accomplishments that we're proud of
I'm proud of all the pages in the app. The Multivariate Clinical Projection, which determines whether a patient’s symptoms are stabilizing or worsening based on complex health markers in recent data, was a really satisfying feature to work on, after working so hard on setting up the Supabase schema. Creating the Care Calendar was also a favorite section of mine, since calendars are such an intuitive way to grasp an significant overview of historical information at once, and this feature allows caregivers to spot emergency flags and medication gaps at a single glance. I really think these features could be useful for tracking and monitoring Alzheimer's symptoms, in order to improve the quality of care and the wellbeing of the patient and their family.
What we learned
I learned a lot about how your approach to building an app really matters in how each aspect and feature is planned and implemented. My initial idea was inspired by a previous project I worked on, related to classifying Alzheimer's severity based on MRI scans. Treatment and care after the moderate case is often limited so care and empathy are really necessary in this process. I also thought that targeting those who are at risk of developing Alzheimer's would allow me to help in their preventative treatment and lifestyle monitoring. Seeing the needs from the caregivers' or users' perspective helped shape the vision and features implemented for the MVP. Mobile responsiveness and user-experience were also critical aspects, since the intended users of the app (caregivers) would usually be multi-tasking, on-the-go, or in moderate to high-stress situations.
What's next for AlzheimerFairy Care System
Several other features could be planned and implemented, but before that I think refining the metrics and fields for data collected would be the top priority. Especially because your data quality often limits or defines what your project is capable of, or what your app would be justified in claiming (in regards to technical responsibility). Modifying how the PDF report is shown, perhaps designing it to be more clinical, is also a consideration. And adding a contact portal or sharing features for the health data through secured channels to, for example, the patient's healthcare provider, would be another feature to consider.
Built With
- flask
- gunicorn
- lucide-react
- next.js
- numpy
- pandas
- python
- random-forest
- render
- scikit-learn
- supabase
- tailwindcss
- typescript
- vercel
