eliran89c/self-hosted-llm-on-eks

This project functions primarily as a deployment guide or boilerplate for running LLMs on AWS EKS. With only 17 stars and 8 forks over nearly two years, it has failed to gain any significant community traction or 'gravity.' Its defensibility is near zero because it relies on standard infrastructure patterns that are now better served by official sources. AWS itself provides robust 'EKS Blueprints' for AI/ML workloads, and tools like SkyPilot or KubeRay offer much more sophisticated abstraction layers for the same task. The high frontier risk and platform domination risk stem from the fact that cloud providers (AWS Bedrock/SageMaker) and frontier labs are simplifying the 'self-hosting' problem through managed services or optimized, vendor-supported containers (like NVIDIA NIMs). This repo is a snapshot of an early 2023 workflow that has since been commoditized by official platform capabilities.