Lessons learnt from creating platforms on Kubernetes
04-17, 10:10–10:40 (Europe/Amsterdam), The Suite

Would you ask a gardener to fix your leaking toilet? Probably not. Then, why would you ask a data scientist to fix a crashing kubernetes pod? Or why would you ask your developers to understand the internals of the PyTorch framework?

Data science is hard enough without also needing to know about Kubernete. Kubernetes application development is hard enough too without having to do data science! Learn about why you can build a platform on Kubernetes that hides Kubernetes complexities from data science users and data science complexities from application developers.

Salman & Mauricio will take you through the painful lessons that were learnt while building such a platform and the struggles faced by data scientists in cloud native hyperspace. Ending with a demo of how they integrated Kubeflow, KFServing, ArgoCD, Crossplane, Dapr and Knative to build an effective platform!


There is no such thing as a single use case platform. Platforms are complex that need to enable different teams with different and disparate skill levels to be more efficient while delivering value. This presentation shows tools that can be used together to provide the foundation to build flexible platforms that can target different use cases and personas without limiting the tools that they are currently using.

This presentation focuses on showing shared challenges between data scientist and application development teams on their journey to run their workloads on top of Kubernetes. Tools like Kubeflow which aggregate other CNCF projects have already gone a long way at integrating general purpose tools for machine learning workflows. On the other hand tools like Crossplane allows us to consume cloud resources using a declarative approach while also simplifying how these resources can be consumed by application teams.

Salman works as an MLOps Engineer at Appvia and a Kuberenetes Instructor at Learnk8s. He has worked with a number of organisations in setting up Machine Learning platforms for teams to operate at scale. He co-founded Cloud Native Wales community. You can also find him on YouTube as Soulman Iqbal where he tries to explain cloud native concepts by simplifying them.

This speaker also appears in:

Knative Steering Committee member for the Knative Project, Knative Functions Co-Lead, and working on the Dapr.io project at @Diagrid. Manning Book Author: Continuous Delivery for Kubernetes. Previously I worked at VMware and Red Hat building tools to help developers be more productive.