Clojure vs Sepsis: path to real time enterprise data science
summary
Real system: SPOT runs real‑time sepsis detection at HCA hospitals using Clojure and Datomic, ingesting HL7 into Kafka and building patient snapshots on Kafka compacted topics. Algorithm reduces simulated screens from ~20,000 to ~700 and raises positive detection to ~95%; deployed to 165 hospitals with measured increases in timely treatment and drops in severe sepsis mortality. Engineering specifics: K-map for parallel Kafka processing, just‑in‑time Datomic rollover, OpenShift + a Jupyter iClojure kernel for remote debugging.
AI-written summary · about
chapters (14)
- 0:00Introduction to Spot
- 0:49Healthcare Technology Context
- 2:13The Sepsis Challenge
- 5:33Developing the Solution
- 7:27Implementing SPOT
- 8:45Technical Hurdles
- 10:32Architecture and Workflow
- 13:14Deployment Results
- 14:11Real-world Impact
- 15:54K-map Development
- 17:10Datomic Challenges
- 19:18Tooling and Open Source
- 21:17Demo and Future Work
- 24:16Conclusion
related talks
18mPipelines for Clinical AI: Real-Time Sepsis Detection
57mLessons in Building a System that Processes More than 70 Billion Events Daily
44mData Science in Clojure
38mComposable Healthcare
25mLifesaving AI and JavaScript
35mSequence and Concurrent Processing for Huge Genomic Data
18mClojure Versus the Botnets
47mReal-time Fraud Detection Challenges and Solutions