Can you deploy DeepSeek-R1 on Kubernetes? In this video, Sai is joined by AWS Solutions Architects Lucas and Tiago to talk about why DeepSeek is different and show you a demo of deploying it on Amazon EKS Auto Mode. We cover concepts like distillation, reinforcement learning, and show how it can run on EC2 instances powered by AWS Inferentia chips and NVIDIA GPUs. We use Amazon EKS Auto Mode for flexibility and scalability on the underlying compute, removing the need for you to manage infrastructure tasks.
Hosting DeepSeek Guide: https://bit.ly/deepseek-tutorial
GitHub Sample: https://bit.ly/deepseek-eks-github
0:00 Intro
1:05 DeepSeek Demo
5:43 Distillation
6:40 Efficiency - R1 Review
11:16 Reinforcement Learning
12:13 Why Kubernetes?
15:39 DEMO: Infrastructure Deployment
24:53 DEMO: DeepSeek on EKS + EC2
31:24 Conclusion