Inspiration
Every second counts during a medical emergency, yet millions of people hesitate because they don't know what to do before professional help arrives. I was inspired to bridge that gap by building an AI-powered first response companion that provides immediate, reliable, and personalized first-aid guidance. My goal is to make life-saving information more accessible, empowering anyone to respond confidently while waiting for medical professionals.
What it does
LoidAI is an AI-powered first response companion that assists users during medical emergencies. Users can describe symptoms or upload an image of an injury, and LoidAI analyzes the situation using AI and Retrieval-Augmented Generation (RAG) to provide accurate first-aid guidance. It delivers step-by-step instructions, educational videos, and assesses the severity of the situation.
For users with supported smartwatches, LoidAI also incorporates real-time health metrics such as heart rate to provide more personalized recommendations. When necessary, it can recommend seeking immediate medical attention and direct users to nearby hospitals. LoidAI is designed to support not replace healthcare professionals.
How I built it
I built LoidAI using React Native (Expo) for the mobile application and PHP, MySql for the backend services. I integrated Google Gemini for multimodal AI capabilities, enabling the application to understand both text and images.
To improve the reliability of responses, I implemented Retrieval-Augmented Generation (RAG) with Pinecone, allowing the AI to retrieve trusted first-aid knowledge before generating responses. User data and application services are managed through a backend API, while Google Maps APIs are used to locate nearby healthcare facilities. I designed the architecture to be modular so future features, such as wearable integrations and offline emergency support, can be added easily.
Challenges I ran into
One of my biggest challenges was ensuring that AI responses remained accurate and responsible in medical situations. I addressed this by implementing a RAG pipeline so responses are grounded in trusted medical knowledge instead of relying solely on a language model.
Another challenge was integrating wearable health data, as different smartwatch manufacturers expose health information differently. I also had to carefully design the user experience so emergency guidance remains clear, fast, and easy to follow under stressful conditions.
Accomplishments that I'm proud of
- Built an AI-powered first-aid assistant capable of understanding both text and images.
- Successfully integrated RAG to improve the accuracy and trustworthiness of medical guidance.
- Created a scalable architecture that supports future wearable integrations and emergency features.
- Developed an intuitive mobile experience focused on accessibility and speed during emergencies.
- Combined multiple AI technologies into a single platform with the potential to make a meaningful real-world impact.
What I learned
Building LoidAI taught me that developing AI for healthcare requires much more than integrating a language model. Accuracy, transparency, and user safety must always come first. I learned how Retrieval-Augmented Generation significantly improves response quality, how multimodal AI can enhance emergency assistance, and how thoughtful user experience design is critical when users may be under stress.
What's next for LoidAI
My vision is to evolve LoidAI into a comprehensive AI emergency companion. Upcoming features include:
- Voice-based emergency assistance for hands-free guidance.
- Offline emergency mode for critical first-aid instructions without internet access.
- Expanded smartwatch integration for continuous health monitoring.
- AI-generated emergency reports that can be shared directly with healthcare providers.
- Support for additional languages to improve accessibility.
- Partnerships with healthcare organizations to validate medical content and expand real-world adoption.
Ultimately, I want LoidAI to become a trusted first-response tool that helps people make informed decisions during emergencies while connecting them to professional medical care as quickly as possible.
Log in or sign up for Devpost to join the conversation.