"Transducing Healthcare: Transducers as Clinical Features" by Derek Schatzlein, Jack Mocherman

Data driven insights are key to healthcare technology. HCA Healthcare, as the largest for-profit healthcare company in the US, is in a unique position to transform healthcare. With streaming realtime patient data from 180+ hospitals, data is plentiful — but it is also eclectic and ill-suited to the needs of data science/realtime decision support applications. Enter Sojourner, a streaming feature engineering platform built at HCA entirely in Clojure/Datomic, and backed by Kafka and GKE. Sojourner uses a lambda architecture of Clojure services to produce features for 40 million patient encounters per year, and it drives a wide variety of applications in our hospitals from sepsis detection to data science. This talk will describe the design of the Sojourner system, how it uses the unique properties of Clojure transducers to construct patient transforms that define features, and the technical/organizational challenges of deploying a fleet of Clojure microservices at true enterprise scale.

Recorded Oct 24, 2024 at Clojure/conj 2024 in Alexandria, VA