Home page
Patient Dashboard
Fall Detection
AI Therapy
Skin Cancer Detection
Help Request Center
Staff Dashboard
Authentication

🛡️ SafeGuard – AI That Cares

Built for HackUSF 2025 | AI in Healthcare

💡 Inspiration

In busy hospital environments, nurses are often stretched thin while patients feel isolated and vulnerable. Our team wanted to create a smart, AI-powered platform that acts as both a care assistant and companion, improving safety, mental health, and diagnostic efficiency. We were inspired by the idea of making healthcare more empathetic, proactive, and tech-enabled, especially for underserved or understaffed facilities.

🛡️ SafeGuard Whole Platform: https://frontendhackusf-4yxjhn9cw-dahomitas-projects.vercel.app/

⚙️ What It Does

SafeGuard is a comprehensive AI platform that helps bridge communication between nurses and patients, while offering intelligent health monitoring tools. It includes:

🗣️ 1. AI Voice Companion (seperate repository: link to Frontend)

A virtual therapy bot and emotional support assistant for patients who need someone to talk to.

🔗 VoiceConvo GitHub
🌐 Live Voice Demo

📸 2. Skin Cancer Detection Tool (seperate repository: link to Frontend)

Upload a photo of a skin lesion and receive a prediction using a deep learning classification model.

🔗 Skin_Detect_AI_Model GitHub
🌐 Model Endpoint

🧍‍♂️ 3. Fall Detection System **(main repository: Frontend + Backend)*

Monitors patient activity via camera input and alerts via SMS staff when a fall is detected.

💬 4. Chat Interface + Authentication (main repository: Frontend + Backend)

Patients and nurses can log in via Google and securely message each other. Role-based data handling ensures safe communication.

🛠️ How To Test Our Endpoints

SkinCancerDetection: https://skin-cancer-api.azurewebsites.net/predict POST request, image included
Backend: go to https://fallguardian-api.azurewebsites.net/api-docs to see all endpoints

🛠️ How We Built It

Frontend: React, TailwindCSS, React Router, Vite , JavaScript
Backend: Node.js, Express, Google OAuth (Passport.js) , Python, JavaScript, MongoDB, Swagger
Machine Learning: TensorFlow, OpenCV, custom image classifier , Keras
Deployment: Azure (backend + ML model), Vercel (frontend + AI voice companion), Docker

We used GitHub Projects for planning, Figma for UI prototyping, and Postman for API testing.

🧗 Challenges We Ran Into

Voice AI Integration: Ensuring smooth, natural audio with low latency in-browser.
Fall Detection Setup: Real-time object tracking and model training were hardware-intensive. Took times to deploy and test using Docker and Azure. Maintain CI/CD piplines
Authentication Management: Role-based auth took time to refactor and secure.
Hackathon Time Pressure: Juggling full-stack development with multiple AI features in one weekend was no small feat!

🏅 Accomplishments That We're Proud Of

Built and deployed three fully functional AI features within one platform.
Designed a smooth chat + login system with live Google OAuth.
Delivered production-ready endpoints and clean UI experience.
Worked seamlessly across multiple tech stacks in parallel.

📚 What We Learned

Deployed ML models via Azure App Services for easy scaling.
Gained expertise in OAuth, session control, and user roles.
Enhanced our frontend design and deployment workflow.
Learned how to collaborate effectively across a full-stack, multi-repo project.