The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence) Building Foundational ML Platforms with Kubernetes and Kubeflow with Ali Rodell - #595
34 snips
Oct 17, 2022 
Ali Rodell, Senior Director of Machine Learning Engineering at Capital One, shares insights from his 25+ years in software development. He discusses the modernization of ML infrastructure in finance, highlighting the transition to cloud-based systems and its unique challenges. Ali emphasizes the importance of a robust DevOps team and open-source tools like Kubernetes and Kubeflow. He also tackles complexities in managing distributed computing and the need for strong software engineering practices to ensure stability and reusability in machine learning platforms.
AI Snips
Chapters
Transcript
Episode notes
Cloud's Impact on ML Development
- Cloud computing helps overcome inertia from legacy systems and regulations.
- This allows for rapid innovation in machine learning at Capital One.
Simplifying Data Science Work
- Make data science easy for everyone, regardless of their technical background.
- Provide clear guidelines and easy-to-use tools within a secure and regulated environment.
Software Engineering Fundamentals in ML
- Machine learning infrastructure problems often boil down to fundamental software engineering principles.
- Constraints like memory, CPU, and disk IO still apply, even with layers of virtualization.