MLOps Isn't That Hard: Modular Stack with Open-Source Tools

This talk, presented by Adam Probst, CEO & Co-Creator of ZenML, explores how to build and manage machine learning systems. It addresses the complexities of code, models, and data throughout the machine learning solutions lifecycle, highlighting the dance between processes, practices, and tools. The presentation introduces ZenML as an open-source MLOps framework that stitches together various tools on an abstraction layer, decoupling code from infrastructure to make them exchangeable. This modular approach allows for delayed decisions on specific tools and ensures reproducibility and auditability. The talk demonstrates how ZenML simplifies the process of moving code from a laboratory environment to production by abstracting away infrastructure specifics, enabling easy switching between artifact stores and tool implementations.