MedGPT
Company
Hackathon Project
Role
Lead Architect
Timeline
2023
Accurate anomaly identification in clinical brain scans
Synchronized visual and tabular diagnostics output
Mental health support and clinical diagnostic fallbacks
In short
MedGPT is a clinical AI diagnostic support agent designed for medical hackathons. It processes visual inputs (MRI brain scans) alongside dynamic patient logs to provide highly precise second-opinion diagnostics.
The problem
Analyzing brain MRI scans requires deep expertise, and early stroke signs can be easily missed. Standard medical models are closed off, heavy, and fail to incorporate qualitative patient logs when outputting diagnosis recommendations.
What I did
Built a multimodal Next.js portal. Integrated OpenAI Vision APIs and customized vector models. Programmed secure authentication streams with Supabase and styled a clean, accessible interface using Tailwind CSS.
The result
- Shipped a production-grade medical prototype under extreme hackathon constraints
- Demonstrated high stroke anomaly identification rates in MRI visual inputs
- Delivered an accessible, responsive dashboard for high-stress emergency departments
Technical detailsFor engineers: how it's built and what it runs on
Architecture
- Next.js 14 routing with serverless Next API routes
- Supabase Auth and PostgreSQL storage backplane
- Multi-modal LLM orchestration via OpenAI Vision pipelines
- Tailwind CSS UI optimized for clinic tablets and desktop viewports
Highlights
- AI-powered MRI scan analysis for stroke detection
- Hybrid input processing: brain scans + patient questionnaires
- Real-time diagnostic feedback with high accuracy
- 24/7 mental health and rehabilitation support integration
- Secure user authentication with Supabase
- Responsive, accessible frontend with Next.js and Tailwind CSS
- Cross-platform compatibility: mobile, tablet, desktop
Stack
Finished reviewing?
Back to Portfolio←