Inspiration

I wanted to make cardiovascular research models easier to explore and understand. Their predictions can be difficult to interpret without context, so I combined clinical inputs, research estimates, and a 3D heart reference in one workspace.

What it does

AngioSight lets users enter available clinical features and review cardiovascular model estimates alongside interactive heart anatomy. It is a research and educational prototype, not a diagnostic tool.

How I built it

I built the interface with React and Vite, the API with Python, and the interactive anatomy viewer with Three.js and React Three Fiber. The frontend is hosted on Vercel, and the API is deployed on Northflank.

What I learned

This project taught me how much work sits between a trained model and a useful application: connecting the frontend and API, handling missing inputs, presenting model outputs clearly, and making a large 3D asset load reliably in a browser.

Challenges

The hardest parts were preparing the models for API use, deploying the service with limited resources, and integrating a detailed heart model without slowing the experience. AngioSight is still a work in progress, and I’m continuing to improve the deployment and model integration.