HealthRisk Analytics.
A hackathon project that turned patient records into a risk classification model and a dashboard for clinical review.
- Model accuracy
- 92.4%
- Records processed
- 10,000+
- Hackathon result
- 1st place
- Delivery
- Production demo
The challenge
Changes in heart rate, oxygen saturation and blood pressure can be difficult to interpret across large sets of patient records. The project explored how a model could help clinicians review several signals together.
The build
We cleaned and standardized patient telemetry in Python using NumPy and Pandas, then trained a classifier to assign risk tiers based on patterns in those measurements.
The dashboard displayed those tiers alongside symptom information so a reviewer could inspect the inputs and results in one place.
Technical deliverables
- Python telemetry pipeline
- NumPy & Pandas
- Risk classification
- REST API
- Responsive clinical interface