Inspiration
Many cultures embrace the wisdom of seeking healthcare judiciously, avoiding unnecessary hospital visits for minor issues while also recognizing the challenges of accessing simple medications without incurring significant costs, given the limitations of the current healthcare system. This limitation gave birth to the concept of Remedi, a harmonious blend of holistic and affordable healthcare solutions, designed to bridge the gap and provide a more balanced approach to wellness.
What it does
Envision a scenario where you are unsure whether your current health problem is severe enough to visit the hospital for, or whether you're being too judicious and risking your precious health. Remedi will provide the perfect solution for free and give you the most holistic opinion on your condition so that you don't have to overspend. Here's our tech features:
Nested Image Classification Machine Learning Models:
A user can upload an image of their problem and our nested ML models use computer vision and 4 classification models to deduce the type of disease, whether it's an eye/skin/ear/oral problem.
Custom fine-tuned GPT Assistant LLM
Based on the text input given by the user, our custom GPT assistant creates an iterative process of asking a few questions about the problem and then deduces a possible cause. Furthermore, it decides a severity level to alert the user on whether they need to visit a doctor.
Community-Service Doctor Appointments
Doctors near you can sign up voluntarily to remedi for community-service such that if our assistant deduces that your case is too severe, you could consult a doctor for free. This feature is catered towards underpriveleged people in the neighbourhood.
How we built it
- Swift & SwiftUI for the front-end
- Appointy for scheduling services
- FirebaseAuth & Firestore for Database and User Management System
- Tensorflow & Keras for Image Classification Model
- AWS S3 Bucket to store user image data
- OpenAI Assistants API for Custom LLM
- Flask + Python for web server and back-end
- Ngrok for web hosting
Challenges we ran into
- Integrating assistants API to make the diagnosis an iterable process
- Connecting the frontend with API calls with specific query parameters
Accomplishments that we're proud of
- Great design choices for the UI
- Intuitive User Experience
What we learned
- Utilizing Assistants API
- Building custom ML Models
What's next for Remedi
- Utilizing more niche LLMs such as Google's MedPaLM
- Training a clustering model with a wider dataset
Log in or sign up for Devpost to join the conversation.