Developer trying to get better at making software and help others by talking and writing about it. Writing tech tutorials, perspectives on problems, tools, culture, devops, business etc. Working with Go and Kubernetes on MLOps at seldon.io
Machine Learning is hot but organisations are struggling to run it in live and MLOps is not easy to master. There’s a dizzying array of tools and they look different from the usual DevOps tools. To apply DevOps skils to ML, we need to understand the specific challenges of ML build-deploy-monitor workflows. We’ll use reference examples to understand the cycle in terms of data prep, training, rollout and monitoring. We’ll see that the key challenges relate to data volumes, changes and quality - challenges alien to the mainstream DevOps world. There are tools to help us but we have to find our way around them.
Getting a better picture of MLOps will help us in lots of ways. It will help us evaluate Machine Learning Operations (MLOps) tools and platforms. We’ll come to understand the specific needs for governance, versioning and monitoring in machine learning. We’ll get a sense for the range of MLOps use-cases and better understand how to scope Machine Learning projects.
Scheduled on Saturday from 13:15 to 14:05 in Stream 2