Doctor Guide

Inspiration

Navigating healthcare can be confusing, especially when you're unsure which specialist to see. We noticed a common pain point: people knowing they're unwell but not knowing which department to visit. Traditional healthcare platforms often fall short here, simply listing services without guiding users through complex symptoms. Our inspiration came from the desire to streamline this initial, often stressful, step.

What it does

Doctor Guide is an AI pre-diagnosis assistant designed to simplify your journey to primary care. You simply describe your symptoms in natural language, and our AI engages in a brief, conversational Q&A. Based on this interaction, it intelligently suggests the most suitable primary care department (like Internal Medicine, Dermatology, or ENT). Beyond recommendations, it automatically generates a concise symptom summary you can use at the hospital reception, saving you valuable time and reducing the need for lengthy explanations.

How we built it

We built Doctor Guide with a focus on speed and efficiency, both in development and user experience. For the frontend, we chose Vite with React to create a responsive chat UI, quickly deploying it using AWS Amplify.

On the backend and infrastructure side, we utilized AWS DynamoDB for our database, interacting with it via the ElectroDB ORM. Our chat API was built with Hono, powered by AWS Bedrock Claude-3 Sonnet for the conversational AI. This entire backend was deployed using Pulumi to AWS Lambda and API Gateway, ensuring a robust REST API communication. Throughout the development process, SwaggerUI was instrumental in facilitating seamless collaboration across the team.

Challenges we ran into

A significant hurdle was building the ETL (Extract, Transform, Load) service. While we successfully built it, we ultimately decided it would be better as a next-step feature. The main difficulty lay in consistently processing public data and inserting it smoothly into DynamoDB. This proved exceptionally challenging. We're currently optimizing this process by loading up to 25 processed data chunks into S3 and using a script to reconfigure and insert any failed data. We received substantial help from Amazon Q during this optimization.

Accomplishments that we're proud of

We're proud to have created a practical AI solution that addresses a genuine user pain point: the uncertainty of initial hospital visits. Building an intuitive conversational AI that accurately guides users to the right primary care department and generates useful summaries, all while adhering to complex medical disclaimers, was a significant achievement. We're also proud of its inherent scalability and localization potential, which we believe makes it highly impactful.

What we learned

We gained significant insights throughout this project. Firstly, we realized the critical need for better localization in healthcare-related services, understanding that medical information and systems vary greatly by region. Secondly, the ETL (Extract, Transform, Load) process proved to be an incredibly challenging and demanding task. And finally, navigating the abstraction layers within AWS presented its own set of complexities. These lessons have been invaluable for our team's growth.

What's next for Doctor Guide

Doctor Guide has a robust future. Beyond its current capabilities, we plan to expand its reach through a dedicated chatbot UI app. We see immense potential for broader application through content localization, adapting it for diverse medical systems and even extending into cosmetic procedure recommendations. The core structure is highly versatile, allowing us to swap medical symptom content for aesthetic concerns (e.g., "fine lines" leading to "skin booster" recommendations). This versatility positions Doctor Guide to become a widely applicable AI assistant for various health-related inquiries, streamlining initial steps in multiple domains globally, especially where geographical distances to specialized care are significant.