Agenda

16:30 Door open, finding seat, enjoying snacks and networking

17:00 Welcome by the Cloud Native Copenhagen organizers

17: 05 Welcome by our host Unity

17:15 Talk 1: The journey of adopting Ray at Unity Ads

18:00 Break with food and drinks

18:45 Talk 2: Optimizing Kubernetes Workload Placement in Multi-Cloud Environments with Project Ultron

19:30 Break with snacks, networking

20:00 Thank you for today

Join us at the Unity Copenhagen office on October 31st for an evening of insights into cutting-edge technologies shaping the future of machine learning and multi-cloud scheduling. This meetup brings together three expert speakers who will dive into the powerful tools that enable scalable and efficient workflows in these domains.

Adam Palmer and Rasmus Selsmark, Senior Machine Learning Engineer, and Staff Software Engineer at Unity, will kick off the event with a talk titled "The journey of adopting ray at Unity applied research".

"Have you heard about Ray, but don't have the time to experiment with it yourself? Experience our journey of adopting Ray for distributed processing and training, over the course of 4 months, condensed into 30 min. Learn about where Ray succeeded and the mistakes we made so you don't have to. This includes scaling from 1 data scientist to 10+ using Ray. In September we ran 4000+ jobs using Ray. This talk will cover how we got there from start to finish."

Next, Tobias Andersen, Freelance, ex Novo Nordisk, will present "Optimizing Kubernetes Workload Placement in Multi-Cloud Environments with Project Ultron" 

"As multi-cloud strategies become the norm for organizations managing complex workloads, Kubernetes is the platform of choice for orchestrating containerized environments. However, multi-cloud setups introduce significant challenges for workload scheduling, such as navigating diverse cloud pricing models, resource configurations, and network performance. Existing tools like Karpenter and Cluster Autoscaler automate node provisioning but lack the sophisticated decision-making needed for optimal workload placement across heterogeneous environments.

In this talk, we introduce Project Ultron, a proposed solution that will enhance Kubernetes scheduling by intelligently labeling pods and nodes in real time. By analyzing cloud provider pricing, compute durability, network characteristics, and node stability, Project Ultron aims to improve resource utilization and lowers operational costs. This session will dive into the complexities of multi-cloud Kubernetes scheduling, discuss how Project Ultron’s algorithm is designed, and explore real-world use cases that optimize workload placement for cost-efficiency, performance, and resilience. Join us to learn how Project Ultron empowers Kubernetes with smarter, more dynamic scheduling for multi-cloud operations."

Don’t miss this opportunity to learn from industry experts, network with peers, and discover practical techniques to enhance your ML and orchestrate workloads across cloud providers! See you there!