The Essentials of MLOps: With Eric Riddoch

Join Miko Pawlikowski as he dives into the world of MLOps with Eric Riddoch, a machine learning platform engineer and MLOps practitioner. In this episode, they discuss the differences between MLOps, DevOps, and platform engineering, tools and practices in MLOps, as well as Eric's journey into the field from studying applied math to becoming an MLOps expert. They explore automated workflows, experiment tracking, model serving, and monitoring, while considering the evolving landscape of MLOps and the challenges of integrating various tools. Tune in for an in-depth look at the technical and non-technical aspects of MLOps, and learn why this field is critical and exciting.

00:00 Introduction to MLOps

01:20 Eric Riddoch's Journey into MLOps

08:12 The Emergence of MLOps

10:23 Comparing MLOps and DevOps

10:53 Challenges in MLOps

21:15 Tools and MLOps Maturity

25:57 Building an ML Platform with Orchestrators

26:35 Experiment Tracking and Model Performance

27:08 ML Flow and Alternatives

29:18 Serving Models with BentoML

31:49 Challenges with SageMaker and GPU Quotas

32:54 Monitoring Tools and Their Limitations

36:48 The PyTorch vs TensorFlow Debate

42:41 Challenges in MLOps Roles and Leadership

50:42 Advice for Aspiring MLOps Engineers