Human-Friendly, Production-Ready Data Science with Metaflow - Ville Tuulos

Explore the groundbreaking open-source framework Metaflow in this 42-minute conference talk by Ville Tuulos, an expert in machine learning infrastructure. Discover how Metaflow empowers data scientists without software engineering backgrounds to create and deploy end-to-end ML workflows. Learn about the framework's user-focused experience, allowing data scientists to concentrate on modeling using their favorite libraries while providing robust solutions for data, compute, orchestration, and versioning. Gain insights from Tuulos' extensive experience, including his leadership of the ML infrastructure team at Netflix. Understand how Metaflow meets data scientists where they are, enabling organizations to build and deploy business-critical ML projects independently across various industries. Dive into topics such as common ML project challenges, the project lifecycle, a simple Metaflow workflow, Metaflow Cards, graceful failure handling, and key lessons learned in this comprehensive exploration of human-friendly, production-ready data science.

- Introductions & Opening
- Recommendations
- What’s Common in ML Projects
- MetaFlow
- Project Lifecycle
- A Simple Metaflow Workflow
- Metaflow Cards
- Handle Failures Gracefully Whenver Possible…
- Lessons Learned
- Curious to learn more?