Avoid ML OOps with ML Ops - A Modular Approach to Scaling End-to-End ML Platforms

Discover a comprehensive approach to building a scalable and efficient Machine Learning Operations (MLOps) platform in this 51-minute conference talk from MLOps World: Machine Learning in Production. Learn how Forethought, an enterprise company developing AI-powered customer experience solutions, transformed their rudimentary ML infrastructure into a mature, modular system. Explore key areas of improvement, including streamlining ML training with Sagemaker, efficient model serving using Sagemaker Serverless and Multi-Model Endpoints, orchestrating ML processes through automated pipelines on Dagster, centralizing feature engineering with Spark, and building intuitive model management tools with Retool. Gain insights into identifying critical components of solid ML infrastructure, addressing bottlenecks, improving ML lifecycle in stages, and implementing best practices for a stable end-to-end ML platform. Understand the journey from a basic v0 architecture to an enhanced v1 system, and get a glimpse of future plans for automated re-training and LLM support in the v2 vision.

Avoid ML OOps with ML Ops: A modular approach to scaling Forethought’s E2E ML Platform