SafeSphere – AI-Powered Digital Safety Companion

Inspiration

The inspiration behind SafeSphere stemmed from a growing concern about the increasing number of people—especially teenagers and non-tech-savvy individuals—falling victim to online scams, unsafe conversations, and privacy breaches. As the digital world expands, so does the complexity of its dangers. We wanted to build a multi-agent AI system that empowers users with real-time protection, guidance, and awareness—all in one platform.

What it does

SafeSphere – AI-Powered Digital Safety Companion is a multi-agent AI system designed to protect users from online threats such as phishing, scam links, unsafe file uploads, and toxic chat environments. It provides: ~virus and scam detection for URLs and files ~Chat safety filtering to identify harmful or unsafe content in conversations ~User authentication and restricted feature access to ensure data protection ~An intuitive React-based frontend that dynamically responds to the user's state

How we built it

We built SafeSphere using a mix of advanced AI tools, web technologies, and cloud services: Frontend: ~React.js with high-level CSS for a modern, responsive design ~React Router v6 for smooth navigation between pages (Home, Login, Signup, Features) ~Conditional rendering and popup prompts for first-time or unauthenticated users

Backend: ~Django REST Framework for APIs and user management ~Virestotal for file scan & url detection ~chat_safety_agent.py for monitoring unsafe conversations ~privacy_score_agent.py for privacy assessment

Cloud & Deployment: ~Vertex AI for model deployment and inference ~Firestore for storing logs and user data ~Google Cloud Storage for secure file uploads

Challenges we ran into

~Synchronizing multiple AI agents efficiently in a single workflow ~Designing a 30–120 second scan delay that wouldn't hurt user experience ~Handling dynamic frontend states and feedback based on auth status ~Creating accurate phishing detection using limited training data ~Time management during the hackathon to build a robust MVP

Accomplishments that we're proud of

~Built a fully functional multi-agent AI system in just a few days ~Integrated Google Cloud tools and ADK into a seamless workflow ~Developed a responsive, modern frontend with smooth user authentication ~Successfully implemented file scan, url scan and chat safety detection ~Designed a UX that balances security, clarity, and accessibility

What we learned

~How to structure and implement multi-agent systems using ADK ~Best practices for using Google Cloud Vertex AI and Firestore ~Tailwind CSS and React Router v6 to create scalable and responsive UIs ~Designing AI tools that are both technically effective and user-friendly ~Importance of secure authentication and feature gating in digital safety apps

What's next for SafeSphere – AI-Powered Digital Safety Companion

~Add the Digital Coach to include voice interaction and smart suggestions ~Add multilingual support for global accessibility ~Partner with schools, parents, and NGOs to promote teen internet safety ~Include real-time alerts and threat history tracking ~Launch as a browser extension or mobile app for seamless daily protection

Built With

• django
• firebase
• google-client-authentication
• huggiface-api
• javascript
• python
• reack
• virestotal-api