AIM214-R1 – Introducing Amazon SageMaker Studio, full IDE for ML (re:Invent 2019) – Key Takeaways

The Key

  • SageMaker Studio overview with live demos

The Takeaways

  • SageMaker is industry’s first ML IDE
  • SageMaker Notebooks = CloudFormation for ML
  • SageMaker Experiments
    • Run a whole ML stack with parameter tuning
    • Store and compare results
  • SageMaker Debugger
    • Zero coding model debugger
    • Analyze your model and training process using rules
  • SageMaker Model Monitor
    • Watches over production model performance
    • Detect drifts
    • Compute feature statistics
  • SageMaker Autopilot
    • Fully automatic ML process
    • User input raw data and select column to predict
    • Algorithm, model creation, training done automatically
    • Transparent notebooks generation

The Non-essentials

  • GE Healthcare case study
    • NOTE: Personally I find this case study a pure marketing for GE Healthcare, no relation whatsoever to SageMaker or SageMaker Studio, thumbs down!
    • AI-assisted ultrasound checking in India
    • Calibrating sonographs automatically


Looks like SageMaker Notebooks is much like CloudFormation for ML