Using R to develop production modeling workflows at Mayo Clinic
summary
A production pipeline predicts which ICU discharges will need the Respiratory Care Unit and runs at Mayo Clinic, using an targets pipeline to avoid full reruns, renv lockfile + git for reproducible environments, and Shiny + plumber API (and vetiver) to serve real‑time predictions via Posit Connect. The codebase is organized as reusable functions so retrospective training and prospective scoring share the same cleaning logic; the system is live and reduces deployment effort without requiring dedicated DevOps.
AI-written summary · about
chapters (6)
- 0:00Introduction to RCU project
- 2:51Overcoming pipeline hurdles
- 7:38Leveraging targets for R
- 10:58Stable production practices
- 14:40Clinical model enablement
- 17:57Q&A session
related talks
17mR & Python playing nice, in Production
29mHow to automate the deployment of ...
11mDesigning Governed AI Pipelines for Healthcare Impact
29mWhen Worlds Collide: Productionalizing a Data Science Model
28mThe COVID Scenario Pipeline: High Stakes Data Science
33mBridging large-scale FHIR to AI in production
18mPipelines for Clinical AI: Real-Time Sepsis Detection
35mResponsible implementation of Machine Learning models for life-and-death decision making