GPUs sitting idle, training environments nobody shut down, pipelines held together by scripts: ML infrastructure gets expensive fast. Can you run a production-grade ML platform on AWS without the cloud bill growing faster than your models? 

Join us for a Svitla Smart Talk with Stanislav Kolenkin, Senior Solution Architect and AWS Community Builder, and Denys Verveiko, Team Lead SRE for the MLOps team at Temabit (Fozzy Group), as they share how to build a scalable, cost-efficient MLOps platform with Amazon EKS and Kubeflow. 

What you’ll learn 

– Architecture – designing scalable ML platforms on EKS and Kubeflow 
– Cost optimization – Spot Instances, Karpenter, and smarter resource allocation 
– GPU efficiency – getting more out of expensive GPUs and cutting idle capacity 
– ML pipelines – automating training, experimentation, and deployment with Kubeflow 
– Observability and FinOps – tracking resource consumption and making ML costs visible 

Key takeaways 

– Why Kubernetes and Kubeflow work well together for ML workloads 
– Practical ways to lower ML infrastructure costs on AWS 
– How to monitor ML workloads and see where the money goes 
– Real-world challenges, architectural decisions, and lessons learned 

Please note that this webinar will be in Ukrainian.   

This event is dedicated to raising funds for an EcoFlow power station and a generator for our teammate in the “Legion North” Unmanned Systems Battalion of the 44th Separate Mechanized Brigade. 

Link to donate https://send.monobank.ua/jar/aj3HKQdS2 
PayPal: d.vladyka@svitla.com 

Register for a practical look at running ML in production without unnecessary cloud spend. 

Speakers 

Stanislav Kolenkin is a Senior Solution Architect with 17+ years in IT. He specializes in AWS, GCP, hybrid and multi-cloud, Kubernetes, containers, and serverless. He is a Kubernaut, Calico Big Cat, and AWS Community Builder (Containers), with extensive experience in mentoring and training.

Denys Verveiko leads the MLOps SRE team at Temabit, where for the past three years he has built reliability practices, first for data platforms and now for ML infrastructure. He brings observability, automation, and production discipline to data and AI workloads. 

Event details: 

Date – 12 November 
Time – UA – 7:00 PM 
Duration – 1.5 hours 
Language – Ukrainian 

All other details will be sent after registration. 
Registration is free and mandatory.