#AWS #AmazonEKS #ArtificialIntelligence #MachineLearning #UltraScale #Infrastructure #Trainium #Accelerators #NVIDIA #GPUs #FoundationModel #Training #Finetuning #Inference #AnthropicClaude #AmazonNova #OpenSource #Kubernetes #etcd #containerd #Karpenter #AmazonVPCCNI #AmazonCSIDrivers #EKSAddons
It took a tiger team of talented passionate builders from multiple domains to get us here. I'm really proud of this bunch and can't wait for what's next! 5/5
These improvements also extend to other AI/ML services built on EKS as their compute layer, like Amazon SageMaker HyperPod with EKS. What's new post - lnkd.in/gJrP3Zwf. 4/n
From gnarly optimizations to architectural improvements to open-source contributions, we made a series of innovations across compute, network and storage to enable frontier model builders like Anthropic Claude and Amazon Nova to run batteries-included accelerated clusters battle-tested for scale 3/n
For the insane engineering behind it, read our tech blog - lnkd.in/g5d8KynK. 2/n
Today marks a quantum leap for Amazon EKS. We're announcing support for ultra scale AI/ML clusters with up to 100,000 nodes or 1.6 million accelerators in a single cluster. Fully Kubernetes conformant at an industry leading scale - lnkd.in/gymjVenN. 🧵1/n
