Open source MLOps at Kubecon with Canonical
Date: 17-21 April 2023
In just a few weeks, Kubecon will be held at RAI Convention Center, in Amsterdam, the Netherlands. After a bunch of news from the industry around AI projects, such as GPT4 or MidJourney4, Canonical is also ready to bring open source into the landscape. Among the attendees, you will be able to meet our machine learning operations (MLOps) experts. Our team is keen to talk about open-source MLOps, demo the latest breakthroughs and answer all your questions about AI.
Open-source MLOps demos at Kubecon
We prepared some fun demos for anyone passing by our booth. Are you curious?
Play StarCraft with Kubeflow
StarCraft is a real-time video strategy game that gamers have come to love for the past 20 years. We are going to demonstrate decision support system (DSS) capabilities within the game. The demo is going to use a closed-loop MLOps pipeline to showcase the influence of reinforcement learning over the gameplay dynamic. It will use Charmed Kubeflow, Canonical’s open-source MLOps platform, which runs on MicroK8s.
The demo will use a map that covers a scenario with a light infantry tactics manual. Within the game, the tactics will change and the units used will depict the light infantry platoon. The model will be trained to figure out what the best reaction is for meeting every Zerg and Protoss unit; 2 of them, 5 of them, and 10 of them. For each of the scenarios, the DSS recommendations and outputs will be presented, showcasing the probability of a win, loss, casualties and more.
For those who prefer a more serious demo, we will also cover how you can secure your machine learning workloads.
Secure MLOps for highly sensitive data
Handling confidential data within an MLOps pipeline can be challenging. This demo depicts a real use case of a genomic-driven data discovery organisation that uses Charmed Kubeflow for its ML workloads. The audience will have the chance to check out how the blood sampling station uses strictly confined MicroK8s and how AppArmor and data tokenization protect Personal Identifiable Information (PIIs). Then the data goes to a local data centre which sometimes needs additional capacity. When such a need arises they use confidential computing machines on a public cloud to host additional K8s workers and scale their MLOps infrastructure in a transparent yet secure way.
This demo is related to the keynote that Maciej Mazur and I will offer on Thursday, 20 April, from 10:05 AM onwards. Join us there to learn more!
MLOps keynote at Kubecon
AI has a lot of applications within different industries, often using highly sensitive data. Together with Maciej Mazur, AI/ML Principal Engineer at Canonical, we will be touching on how to handle this kind of data using secure MLOps. We will focus on the real example of a genomics-driven drug discovery organisation using Kubeflow for its MLOps.
After the talk, you will understand how you can set up a secure foundation for machine learning with open-source building blocks. We will cover how confidential computing on the public cloud helps you address run time insecurity. You will then learn how Kubernetes’ strict confinement helps you get complete isolation, up to a minimum access level to the host resources. Finally, we will cover how tokenization can enable you to avoid data leaks and drive high system productivity at the same time. We will demonstrate how this works in practice with a life sciences use case powered by Charmed Kubeflow.
Meet us at Kubecon EU 2023
What is changing in the AI world? How is AI shaping different industries? Are you curious how Canonical can help you reach your AI objectives?
From scaling AI initiatives to showcasing the latest breakthroughs, Canonical’s MLOps experts would like to meet you. Open source MLOps is connecting the landscape, having Charmed Kubeflow as a foundation for a growing ecosystem that helps organisations move further with their AI projects.
Canonical’s helping people and organisations adopt to cloud-native the Ubuntu way: easy, lean, secure, and cost-effective. Our team at Kubecon will cover a wide range of topics such as Ubuntu Pro, MicroK8s AWS Appliance or VM to container workload migration using Charmed Kubernetes.
Canonical presence at KubeCon + CloudNativeCon Europe 2023Read more
- [Whitepaper] A guide to MLOps
- [Solution brief] Enterprise AI at scale with NVIDIA and Canonical
- [Blog] From data-centric to model-centric MLOps
- Ubuntu AI on Medium
Run Kubeflow anywhere, easily
With Charmed Kubeflow, deployment and operations of Kubeflow are easy for any scenario.
Charmed Kubeflow is a collection of Python operators that define integration of the apps inside Kubeflow, like katib or pipelines-ui.
Use Kubeflow on-prem, desktop, edge, public cloud and multi-cloud.
What is Kubeflow?
Kubeflow makes deployments of Machine Learning workflows on Kubernetes simple, portable and scalable.
Kubeflow is the machine learning toolkit for Kubernetes. It extends Kubernetes ability to run independent and configurable steps, with machine learning specific frameworks and libraries.
The Kubeflow project is dedicated to making deployments of machine learning workflows on Kubernetes simple,
portable and scalable.
You can install Kubeflow on your workstation, local server or public cloud VM. It is easy to install with MicroK8s on any of these environments and can be scaled to high-availability.