Inspiration
Interactions with the Lighthouse Medical Centre, we got to know that getting proper healthcare is often overwhelming, especially for underserved communities like refugees, international students, and elderly individuals with language or accessibility challenges. Moreover, navigating insurance options can be confusing, and wait times for essential healthcare services can lead to frustration and dissatisfaction. We were inspired to use AI to create LightHouse.AI, which can simplify this process by offering a platform that makes healthcare management easier and insurance plan recommendations more accessible and personalized.
What it does
LightHouse.AI streamlines healthcare management by offering a web-based solution that: Recommends personalized health insurance plans. Provides an AI-powered interface for booking appointments and comparing insurance options. Addresses user-specific needs, including language preferences, age-related concerns, and accessibility requirements. Optimizes medical staff task allocation, reducing patient wait times and improving service efficiency.
How we built it
We began by gathering insights through in-depth discussions with mentors, medical staff, and healthcare professionals at Lighthouse to better understand the real-world challenges faced by patients and staff. Based on this input, we designed a user-friendly interface that simplifies healthcare management.
For the recommendation system, we tested multiple industry-leading AI models, including those from Hugging Face, OpenAI, and Meta. After evaluating factors like token limitations and result accuracy, we selected LLaMA through Hugging Face as our final model for its optimal performance.
Collaboration played a key role in development, with team members working on different components simultaneously. GitHub was instrumental in managing version control and integrating all parts seamlessly before final testing.
Challenges we ran into
Selecting the right AI model: With so many options available, we had to test and evaluate multiple models. Some produced low accuracy results, while others had limited free tier tokens or lacked the required customization. Limited domain knowledge: Initially, our understanding of the medical and insurance sectors was limited. We overcame this by consulting mentors and staff, and conducting thorough research. Data availability: Finding datasets that included both health insurance packages and relevant health factors proved to be a challenge.
Accomplishments that we're proud of
Developed a working model capable of recommending personalized health insurance plans based on user needs. Implemented classification techniques that streamline internal team communication, reducing the time it takes for staff to coordinate tasks.
What we learned
The importance of teamwork and supporting each other throughout the project. The value of open discussions—no question is too small to ask. Gaining hands-on experience with AI models and their applications. How to design a customizable, user-friendly UI.
What's next for LightHouse.AI
Implementing service classification for voice-based assistance, directing users to the appropriate department and following up on requests. Expanding our services globally, including weekly checkups, allied health, dermatology, gynecology, dental care, and vaccinations. Introducing automated follow-up processes, allowing for rescheduling in case of missed appointments.
Log in or sign up for Devpost to join the conversation.