LinkCare

Selecting a patient to begin a session with
Sample conversation of supporting ADHD communication
Sample generated Mermaid diagram

Inspiration

A blind patient doesn't need a text summary, they need audio. An ADHD patient doesn't need a paragraph, they need a visual map. An ESL patient doesn't need complicated medical jargon, they need understandable vocabulary. We built LinkCare to bridge the gap between a doctor's expertise and a patient's reality.

What it does

LinkCare is a doctor-facing AI tool that uses Anthropic’s Claude Sonnet 4.5 to transform complex medical notes into the exact format a patient needs:

For Visual Thinkers (ADHD): Breaks linear instructions into generative visual flowcharts (using Mermaid.js) to map out routines.

For Auditory Learners (Visually Impaired): Replaces visual metaphors with tactile descriptions and reads them aloud via native Text-to-Speech.

For ESL Learners: Flattens complex syntax to CEFR A2 levels while maintaining clinical accuracy.

How we built it

AI Core: Anthropic Claude Sonnet 4.5.

Backend: Python FastAPI.

Frontend: React + Vite + Tailwind CSS.

Visuals: Mermaid.js.

State: Zustand.

Challenges & Accomplishments

The biggest challenge was creating the multimodal consistency, which was getting Claude to reliably generate valid code(for diagrams and natural language in a single response. We solved this with a robust protocol that ensures the app never crashes, even if the API fails.