ML Platform for Insurance Conglomerate
summary
A cloud-native ML platform is built for a major US insurance conglomerate, migrating a legacy on‑prem stack to AWS and automating CI/CD so model delivery time collapses; model count scales from ~30 to over 100. Core pieces: a feature catalog of feature packages (Python functions reused in dev and production), batch processing on EMR with PySpark orchestrated by Airflow, serving on SageMaker, and experiment/version tracking with MLflow; development environments use obfuscated production-like data and the platform works in production.
AI-written summary · about
chapters (8)
- 0:00Insurance ML platform intro
- 1:17Legacy platform challenges
- 4:33Cloud migration strategy
- 9:12Key technical solutions
- 15:39Cloud architecture details
- 18:15Business and technical benefits
- 21:11Dev environment optimization
- 23:21Q&A
related talks
26mManaging ML Models @ Scale
27mML Platform Journey
31mUse-Case Driven ML Platforms
18mArchitecting Intelligent Platforms
29mBuilding ML Platform Capabilities at Global Scale
32mTeam Teardown: Evolving an ML Platform
36mEnterprise-grade Full Stack ML Platform
28mBuilding Developer-Centric ML Inference Platforms