Prescripto

Clinician Dashboard
Patient Profile
Adding Visits
Notes Canvas for Clinicians
Approving/Denying AI Suggestions

📖 About the Project

Inspiration

As international college students, navigating doctor visits on our own for the first time—without our parents—we were unprepared for steep, surprise copays and often left scratching our heads at paper prescriptions. That frustration inspired us to build a platform that delivers full cost transparency and AI-powered insights into every medication decision.

What We Learned

Mastered React hooks, TypeScript, and Vite for a polished frontend experience.
Leveraged Supabase real-time channels to synchronize hand-drawn notes via QR-code.
Built a Spring Boot backend to securely orchestrate Google Gemini LLM calls and manage sensitive patient data.
Honed prompt-engineering skills to generate allergy-safe, coverage-aware prescriptions.

How We Built It

Frontend: React + TypeScript + Vite for fast, responsive UI and real-time drawing sync.
Backend: Spring Boot service centralizing LLM prompts, enforcing business rules, and exposing a clean automatic LLM response verification endpoint.
Data Store: Supabase/Postgres for patient records, visits, prescriptions, and drawing updates.
AI Integration: Google Generative AI (Gemini) for OCR, prescription recommendations, post-prescription chatbot and copay extraction from insurance formularies.

Challenges

Real-Time Drawing Sync: Implementing seamless QR-code based note capture over Supabase channels.
Data Parsing: Extracting structured allergy and insurance details from diverse, unstructured records and constructing PDF prescriptions.
Prompt Engineering: Crafting robust prompts that avoid unsafe suggestions and respect coverage constraints.
Security: Ensuring API keys and PHI stay secure by routing LLM calls through our backend (we have a lot of API keys...)