R & Python playing nice, in Production
summary
Continental builds a production ML pipeline that replaces ~4 months of physical tire testing with overnight batch predictions and serves over 100 developers daily; the proof‑of‑concept has 1,000+ commits and an 18‑month history. Data moves between steps as Parquet files and company access uses PostgreSQL. Environments use Renv, Poetry, and Docker with separate language images; orchestration runs on Valohai/Kubernetes, EC2 and S3, and outputs feed a Tableau dashboard.
AI-written summary · about
chapters (7)
- 0:00Introduction
- 2:57Team changes to Python
- 6:01Data exchange rules
- 8:30Reliable deployment
- 11:52Multilingual pipeline
- 14:23Final deployment impact
- 15:37Q&A
related talks
14mRobust R Deployments: Building a Pipeline from RStudio to Production
17mProductionizing Data Science Workloads
29mMachine learning in production
40mBuilding highload ML powered service
33mHow you really get your data science models into production the cool way!
21mUsing R to develop production modeling workflows at Mayo Clinic
44mML in Production
25mContinuous ML Applications in Production