Productionizing R workloads using Amazon SageMaker, featuring Siemens
summary
R code runs inside custom Docker images on SageMaker (pushed to ECR) using reticulate and the SageMaker SDK to run SageMaker processing, training, and model registry steps triggered by EventBridge and orchestrated with Step Functions. A weekly pipeline compares new feature-store data, automatically retrains and evaluates models (including ARIMA, ETS, Prophet), sends an approval email, and registers winners; Siemens Energy scales forecasting to 22,000 models over 4000 time series and can spin up to 100K training jobs in parallel.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 0:24Agenda
- 1:40Overview
- 2:37Users
- 3:40Scale
- 7:11Demo
- 7:36Prototyping
- 14:06Reflection
- 14:52Siemens Energy
- 16:56Use cases
- 19:33Detailed use case
- 21:59Architecture
- 23:41Training workflow
- 25:30SageMaker integration
- 27:21Summary
related talks
19mUsing Amazon SageMaker to develop and deploy ML models
12mLeveraging AWS and Kubernetes for Scalable AI/ML Workloads in the Cloud
33mBuild, train, and deploy ML models using Amazon SageMaker
19mDrive transformation through machine learning
19mDeveloping large-scale machine learning algorithms on Amazon SageMaker
34mImplement MLOps practices with Amazon SageMaker
20mIndustrial-Scale Machine Learning
39mTrain ML models at scale with Amazon SageMaker Training