UHackathon • 2nd Place
UHEALTH
UHealth is a deep learning prototype that predicts pneumonia from chest X-rays. Built under a six-hour hackathon deadline, the project combined model development, interface design, and deployment best practices into an easy-to-understand project.
- • Convolutional Neural Network with 92% accuracy (pneumonia vs. normal)
- • Extended ResNetv2 to 4 respiratory diseases with 95% accuracy
- • Cleaned and augmented 10k+ image dataset for robust training
- • 2nd place in Tacoma's largest hackathon ever
From hackathon demo to deployable system
The project began as a binary pneumonia detector but scaled to a multi-disease classifier with a 95% accuracy rate. An automated MLOps pipeline enables ongoing retraining and seamless deployment, demonstrating how even a short-term prototype can lay groundwork for a production-ready tool.
- • Result speed: 3–5 seconds on mid-range laptop hardware
- • Automated CI/CD ensures model freshness and reproducibility
- • Judges cited UX clarity as a key strength — 2nd place overall