Challenge

UCSF wanted to create a scalable model to accurately predict the number of infections and impact of different public health interventions (mask ordinances, lockdowns, etc.) in any particular neighborhood within 7, 14, or 30 days.

Solution

The IKASI AI-based platform created a series of simultaneous models tailored for different geographies and COVID conditions. The models were flexible to include updates from the epidemiologists at UCSF.

Results

■  IKASI created highly accurate predictive models that cut the error rates by more than half:  +/- 3% compared to industry models (coming from academic labs), which had high errors of +/- 8%.

■  The models were “productionized” within two weeks which was 7x faster than the industry standards.