This talk discusses the Autodesk Machine Learning Platform built with Metaflow. It serves as the cornerstone of its managed training infrastructure. It explores the initial integration stages, highlighting the seamless connection between Metaflow and Sagemaker Studio for training job initiation and Metaflow UI access. It dives deep into the mechanisms behind enabling distributed training in Metaflow, the strategic incorporation of GitOps for efficient workflow orchestration, use of other managed AWS services like FSx filesystem and EFA, along with various other enhancements that strengthen the training framework.
Discover more such stories at slack.outerbounds.co
