CareCanvas — The agentic layer for modern care

Overview

CareCanvas turns doctors’ natural handwriting into structured data and continuous patient guidance. We’re building a three-agent system that captures notes, explains them to patients, and closes the loop with analytics and proactive care.

What we’re building at the hackathon

• Clinical Scribe Agent ( live )

  • Doctors write on an 8×11″ digital canvas using digitilizer pen on their regular notepad (Fabric.js).
  • Gemini 2.0 Flash converts handwriting to structured medical JSON (vitals, symptoms, diagnosis, meds) and saves to Firestore with the canvas snapshot.

• Patient Intelligence Agent (in progress)

  • Pulls each patient’s notes/history to power a chat UI.

  • Users can upload Current and Past Blood Reports to visualise them and see the trends of their biomarkers.

  • Answers “What did my doctor say about X?”, explains lab PDFs, and renders trends (Chart.js).

  • Automates nudges: refill reminders, follow-up prompts, progress milestones.

  • Users can connect their wearables data via Health Connect for Android and Apple Health for IOS.

Architecture
React + Vite frontend, Fabric.js canvas, Firebase (Auth/Firestore/Storage/Functions), Gemini 2.0 Flash for OCR + chat, LangChain for tool orchestration and memory, vector DB for retrieval, HIPAA-ready deployment with BAA and fine-grained Firestore rules.

User journey

1. Doctor writes as usual for an X Patient Session → Scribe Agent commits structured note.

2. Patient opens CareCanvas → asks questions (“Why am I on this med?”) → gets plain-English answers with optional trend charts.

3. Proactive automations trigger refills/follow-ups; clinicians review concise summaries and population insights.

Why this matters

• Time back to clinicians: Replace manual charting with instant structure.

• Patient clarity 24/7: Turn jargon and labs into understandable guidance.

• Proactive care: Identify risks and care gaps; improve adherence via reminders.