Pipelines for Clinical AI: Real-Time Sepsis Detection
summary
Claims a pipeline that ingests real‑time vitals and labs, runs feature engineering and inference with Apache Kafka and geo‑redundant microservices, and targets 99.95% uptime with subsecond alert latency. Implementation shows phased rollout: retrospective validation, shadow deployment, canary release, then full production with continuous model drift detection and automated retraining. Observability focuses on sensitivity/specificity, end‑to‑end latency, data‑quality gates and audit trails tied to clinical action.
AI-written summary · about
chapters (10)
- 0:00Introduction and Speaker Background
- 0:19The Challenge of Sepsis
- 1:50Importance of Timely Data and DevOps
- 2:23Complexities in Implementing DevOps for Healthcare
- 3:48Platform Engineering Solutions
- 7:43Streamlining Data Pipelines
- 11:00Phases of Implementing Clinical AI
- 12:00Observability and Data Quality
- 13:53SRE Practices for Healthcare
- 16:52Conclusion and Key Takeaways
related talks
25mLifesaving AI and JavaScript
11mDesigning Governed AI Pipelines for Healthcare Impact
11mEnsuring Reliability in Healthcare Data Engineering
26mEngineering Excellence: Platforms and Architectures for Real-Time AI Analytics
29mAI-Powered Anomaly Detection in Cloud Data Pipelines
28mBuilding a Coud Native Data Pipeline for Your Real-Time ML & AI Apps
26mClojure vs Sepsis: path to real time enterprise data science
16mOperationalizing AI in Finance: Real-Time Analytics