CareSight Landing Page – your gateway to connected care.
Care Mode – Clarity, comfort, and quick access to care.
Document Scanner – Upload and instantly analyze medical records with AI-powered insights.
Dashboard – A personalized overview of your health, alerts, and next steps.
Ask Health Questions – Speak naturally, and CareSight will listen, understand, and respond.

🛠️ Project Story

💡 Inspiration

CareSight was born out of a deep realization: while digital health solutions are rapidly advancing, many vulnerable users—especially the elderly, disabled, and neurodiverse—remain underserved. Medical forms are filled with jargon, interfaces are overwhelming, and voice interactions are rarely prioritized. We wanted to humanize medical technology and create a companion that doesn’t just deliver information, but also translates, speaks, and cares.

We imagined a platform where:

A grandmother could hear what her prescription means in plain language.
A caregiver could be instantly notified if something goes wrong.
A user with motor limitations could navigate everything by voice or keyboard alone.

That’s how CareSight was envisioned—an AI-powered assistant that reads, explains, and speaks health, like a companion, not a cold interface.

🚀What We Built

CareSight is a full-stack AI healthcare accessibility platform powered by Next.js, AI models, and custom accessibility tooling. Key innovations include:

📄 Document Simplifier: Upload a discharge summary or prescription, and our AI reads and rewrites it in plain English (OCR + NLP pipeline).
🗣️ Voice Everywhere: From chat to summaries, everything can be read aloud using the Web Speech API.
🤖 Medical Q&A Chatbot: Ask anything—symptoms, medication explanations, or “what does this mean?”—and get human-style answers from Google Gemini.
🎨 Care Mode UI: A fully adaptive mode with calming colors, large spacing, and empathetic interactions.
🧠 Health Monitoring (Beta): A real-time posture/tremor analyzer using MediaPipe, with optional alerts and privacy-focused local processing.
📬 Caregiver Alerts (Coming Soon): One-click caregiver notifications via EmailJS and SMS options
🌐 Multilingual Support: Integrated translation-ready design, powered by AI-backed summaries.

🧗‍♀️ Challenges We Faced

🤖 Gemini Integration Pains Integrating Google Gemini Pro for medical Q&A was trickier than expected. Early attempts resulted in overly cautious responses like “please consult your doctor.” We had to refine our prompts, safety layers, and fallback strategies to make the assistant genuinely helpful while still ethical and non-diagnostic.
📋 Meaningful Text Extraction While Tesseract.js handled OCR decently, getting it to not just extract—but intelligently organize—medical data (e.g., patient name, doctor, date) into a user-friendly sidebar layout required extra custom logic. Automating that without backend parsing was a major hurdle.
🫀 The 3D Heart Model Battle We attempted to add a calming, educational 3D heart model—but embedding it interactively inside the interface consistently broke or clashed with our layout. Between performance issues and weird canvas conflicts, it was a last-minute decision to either heavily tone it down or drop it temporarily.
📦 UI Data Placement Was Tough Having the user input optional details (like symptoms, conditions, etc.) and then dynamically reflecting that in the right spots throughout the app (chat, summary, alerts) without breaking the layout was harder than it sounds. It involved careful prop management, state lifting, and clean data flow across pages.
⌛ Time, Always Time. As with every hackathon, scope and time were our constant rivals. We aimed high, so we had to make strategic trade-offs, modularize our work, and focus on building a polished MVP while leaving advanced features like SMS, email alerts, and AI summaries for the future.

Deployment Drama

Every time we fixed something, deployment broke something else. We ran into endless Vercel build errors, environment mismatches, and caching issues. Even small UI changes sometimes triggered full rebuild failures or lost connections to the API keys.

Made with:

Frontend- React.js, Next.js, TailwindCSS, Framer Motion OCR Engine- Tesseract.js AI Models- Google Gemini Pro (chatbot + summarizer) Text-to-Speech- Web Speech API, React Speech Kit Email Alerts- EmailJS SMS (Planned)- Twilio Health Monitoring- MediaPipe (Posture, Gesture, Facial Analysis) Authentication- Clerk