Productizing Machine Learning over Big Data with AWS tools
summary
Demo shows how to go from notebook prototype to a deployable ML product using SageMaker notebooks and training jobs, scikit‑learn locally and EMR Spark distributed training for scale, with pipelines that preserve feature transforms. Uses Boston housing data, gets similar R2 on scikit‑learn and Spark, uploads CSVs to S3, runs batch transforms and creates real‑time endpoints and batch transforms (endpoint created but not invoked live in demo). Talks limits: fragile assumptions, CI/CD, metrics, data versioning and A/B testing pitfalls.
AI-written summary · about
chapters (4)
related talks
34mDeploying a Machine Learning System That Only Runs Locally
01h 41mGetting started with Machine Learning on AWS.
44mPragmatic Machine Learning in the Cloud
52mPreparing data for reporting, ML, and AI on AWS
17mProductionizing Data Science Workloads
51mSpotlight on Emerging Data Science Worskspaces and Spark as Development Levers
23mRethinking scalable machine learning with Spark ecosystem
25mDesigning an AWS Well Architected Framework for AI/ML