Inspiration

After a hospital visit, patients are often sent home with dense discharge paperwork filled with clinical terminology. From personal experiences and conversations with family members, we noticed how easy it is to misunderstand medication timing, dosage, or recovery instructions. These misunderstandings can lead to serious health risks, readmissions, or delayed recovery. We wanted to build something that bridges the gap between medical expertise and patient understanding, making recovery safer and more accessible.

What it does

Recoverly is a program that allows people who were just discharged from a hospital or a doctor's checkup to upload their discharge paperwork to create personalized recovery plans to ensure that patients recover safely and smoothly at home rather than leaving them to interpret dense medical languages on their own. It is a system that creates an interactive recovery roadmap with adaptive guidance and risk aware monitoring.

How we built it

We built Recoverly as a web app using Next.js. Users upload discharge paperwork, which is processed using AI to extract key medical information like medications, tasks, and recovery steps. This structured data is then stored and managed so it can be displayed back to the user as a personalized recovery plan. The frontend displays this information in a clean, interactive dashboard that shows recovery progress and key actions. We organized the system into separate parts for the user interface, AI processing, and data handling to keep the project structured and easy to build during the hackathon.

Challenges we ran into

Handling messy medical documents was hard: Discharge papers come in many different formats, so we had to carefully guide the AI to correctly pull out important details like medications, dosages, and follow-up dates without missing anything. Building the Recovery Confidence Score was tricky: We made a score (0–100) that combines symptoms, medication use, and task completion. Getting it to respond properly took a lot of tweaking. Connecting everything was challenging: Putting together user login, live updates, and the AI processing in one system caused bugs that we had to quickly find and fix during the hackathon. TypeScript slowed us down at first but helped later: It took extra time to set up shared data types, but it helped prevent a lot of bugs as the project got bigger.

Accomplishments that we're proud of

Built a system that takes a patient’s discharge PDF and turns it into a clear, easy-to-follow recovery plan during the hackathon.
Created a Recovery Confidence Score that changes based on how well someone is following their recovery plan, so they can quickly see their progress.
Added warning features that alert users about serious symptoms they shouldn’t ignore to help keep them safe.

What we learned

Getting the AI to pull out medical info correctly was harder than expected as small changes in how we wrote prompts could change the results a lot.
Designing for healthcare is different from normal apps because we had to keep everything very simple and easy to understand, especially for older users.
Using Next.js and Supabase was powerful, but it took time to figure out how to properly manage different parts of the app (server vs. client).
Using TypeScript across the project helped us catch bugs early and made it easier to scale the code.
We learned that if the AI isn’t fully sure about something, it’s better to show that uncertainty instead of risking wrong medical info.

What's next for Recoverly

In the future, we would like to expand Recoverly to make it more accessible and impactful by:

Add multilingual support for non-English speakers
Simplify the interface for easier navigation
Integrate voice assistance for hands free guidance
Enable image scanning for physical prescriptions and documents
Partner with healthcare systems like Johns Hopkins Medicine to integrate directly into patient portals
Implement stronger safety features such as verification layers and clinician review option