Dawini (Ψ―Ψ§ΩΩΩΩ) is an AI-powered mobile medical assistant application designed to bridge the gap in accessible healthcare for Arabic speakers. As our graduation project, it delivers a seamless user experience on mobile devices, guiding users from initial symptom understanding to appropriate specialist recommendations, enhancing health literacy and access to care.
Developed with Flutter for the robust mobile interface, alongside Firebase, Flask, and custom-trained AI models deployed on Lightning Studio.
Millions of Arabic speakers lack access to reliable, personalized medical guidance in their native language. They struggle with:
- Accurately describing symptoms.
- Navigating overwhelming and often unreliable online health information.
- Identifying the correct medical specialty for their condition.
Dawini addresses these critical gaps by providing an intelligent, culturally relevant digital health companion accessible directly on a mobile device.
Our system leverages two main AI models for a comprehensive approach, integrated within a user-friendly mobile application:
-
Conversational Symptom Analysis (LLM):
- Purpose: Allows users to describe symptoms naturally in Arabic via a chatbot interface within the app.
- Model: A fine-tuned Aya Expanse 8B (Large Language Model) to ensure context-aware and medically relevant responses. We significantly refined its prompt structure for improved accuracy and tone.
-
Medical Specialty Classification:
- Purpose: Based on structured user input collected through the app's forms, accurately predicts the most suitable medical specialty for referral.
- Data Foundation: We extensively preprocessed a large Arabic Healthcare Dataset, carefully merging and balancing categories into 20 distinct medical specialties for robust classification.
- Model: A robust Hybrid CNN-BiLSTM architecture.
- Intuitive Mobile User Interface (Flutter): Our primary focus was building a clean, responsive, and easy-to-navigate mobile application to ensure a seamless user experience.
- AI-Powered Chatbot: Natural language symptom description and intelligent responses directly within the app.
- Intelligent Triage: Accurately classifies user symptoms into specific medical specialties via structured forms.
- Personalized Doctor Recommendations: Filters doctors by predicted specialty and user's GPS location, displayed clearly on the app.
- Automated Medical Reports: Generates structured PDF summaries of user cases, securely stored and shareable with doctors through the app.
- Dual User Roles: Separate, tailored interfaces for Patients and Doctors, managed through the mobile application.
- Arabic Language Support: Designed from the ground up to support and understand Arabic medical dialogue.
Explore Dawini's user interface through key screens from both patient and doctor flows:
| Area | Technology |
|---|---|
| Mobile App (Frontend) | Flutter |
| Backend & API | Flask, FastAPI |
| Authentication | Firebase Authentication |
| AI Model Deployment | Lightning Studio |
| Data & Reports Storage | Firebase Firestore, Cloudinary |
- π Download APK, Demo Videos,Single Screenshots, and Final Report
- π₯οΈ View Our Presentation Slides
Note on Live AI Demo:
While our AI models were successfully deployed for development and testing on Lightning Studio, public real-time access might be limited by free-tier usage constraints. We are actively working on migrating to a more sustainable hosting solution for continuous availability of the full AI features. The core application logic and user interface are, however, fully functional.
dawini/
π View Full the code files outline
- π Academic Year: 2024β2025
- π« Graduation Project β Artificial Intelligence Department, Cairo University