Exercise: Create a site or blog from your GitHub repositories with GitHub Pages

This is an opinionated umbrella Helm chart for your local single-node **k3s + NVIDIA GPU + Ollama + Open WebUI + LangChain/LangSmith** setup.

CUDA-first OpenTelemetry Python SDK for LLM inference observability and explainability.

Kaggle-Dropbox-HuggingFace

Pre-built llama.cpp CUDA binary for Ubuntu 22.04. No compilation required - download, extract, and run! Works with llcuda Python package for JupyterLab integration. Tested on GeForce 940M to RTX 4090.

Production-grade C++20/CUDA distributed LLM inference system with TCP networking, MPI scheduling, and content-addressed storage. Features comprehensive benchmarking (p50/p95/p99 latencies), epoll a…

Custom CUDA implementation for LLM inference with MPI-based distributed computing. Memory-efficient layer offloading, multi-rank coordination, and GPU optimization for constrained hardware (1GB VRAM).

High-performance TCP inference gateway with epoll async I/O for CUDA-accelerated LLM serving. Binary protocol, connection pooling, streaming responses. Zero dependencies beyond POSIX and CUDA.

CUDA-aware OpenMPI integration for GPU-accelerated distributed computing. Multi-GPU LLM inference with MPI communication, performance benchmarking, and collective operations testing.

Content-addressed LLM model distribution with SHA256 verification and SeaweedFS integration. Distributed storage, manifest management, LRU caching, and integrity checking for GGUF models.

Distributed MPI scheduler with work-stealing algorithm for LLM inference. Percentile latency analysis (p50/p95/p99), throughput benchmarking, multi-rank load balancing, and empirical performance me…

Qt-style CMake superbuild demo: FetchContent deps, feature flags, install/export targets, CI matrix, tests, and CPack packaging.

MCP stdio server for Windsurf that routes tool calls to a local llama.cpp llama-server (GGUF), optimized for low-VRAM GPUs.

This is a sample project to use wolfram with llama.cpp