ForceInjection/AI-fundermentals

Covers GPU/CUDA internals, cluster operations (NCCL, InfiniBand monitoring), and Kubernetes-based AI infrastructure including device plugins, job scheduling with Kueue+HAMi, and GPU virtualization techniques for resource sharing. Combines low-level kernel optimization (TileLang, Tensor Cores) with cloud-native abstractions—addressing the full stack from hardware interconnect (NVLink, GPUDirect) to distributed storage (JuiceFS) and LLM application orchestration frameworks like LangGraph and Spring AI.

914 stars. Actively maintained with 102 commits in the last 30 days.