Highlights
- Pro
Lists (32)
Sort Name ascending (A-Z)
AI Efficiency
Attention
Checkpointing
Collective Communication
Compilers
CUDA
Datasets
Diffusion
Distributed Systems
Graphs
gRPC
Hardware
HPC
Infrastructure as Code
Jax
Job Orchestration
K8s
Kernel Language
Language Models
Linux
Mechanistic Interpretability
NVIDIA GDS: DMA/RDMA
PEFT
PyTorch
Quantization
Research Tools
Rust Machine Learning Ecosystem
Serving
Simulation
Storage
Vision
WASM
- All languages
- Agda
- Assembly
- Bikeshed
- C
- C#
- C++
- CMake
- CSS
- Common Lisp
- Coq
- Cuda
- Cython
- D
- Dockerfile
- Elixir
- Emacs Lisp
- F#
- Fortran
- GLSL
- Go
- HCL
- HTML
- Handlebars
- Haskell
- Java
- JavaScript
- Jinja
- Jsonnet
- Julia
- Jupyter Notebook
- Koka
- Kotlin
- LLVM
- Lean
- Lua
- MATLAB
- MDX
- MLIR
- Makefile
- Markdown
- Metal
- Mojo
- Nim
- OCaml
- Perl
- PowerShell
- Python
- Rocq Prover
- Roff
- Ruby
- Rust
- SCSS
- SMT
- Scala
- Shell
- Starlark
- Svelte
- Swift
- SystemVerilog
- Tcl
- TeX
- TypeScript
- Typst
- Verilog
- Vim Script
- WebAssembly
- Zig
Starred repositories
NVIDIA Kubernetes Engine (NVKE). A management plane for GPU Kubernetes on bare metal.
Towards a More Autonomous Agent System
The lightest sandbox. A process-based sandbox for Linux, no container, no VM, no privilege, no prompt injection
one-year LLM serving trace from Chutes
Secure, fast, efficient, battle-tested at Alibaba's scale. Hybrid architecture code review tool: deterministic pipelines + LLM Agent, precise line-level comments, built-in multi-language ruleset (N…
SoL-Pi: Scaling Auto-Research Loops for Efficient Agent Harnesses
Data and analysis for long-context API TTFT scaling
Lightweight, decentralized KVM/QEMU + LXC orchestrator in Go — a post-Broadcom Proxmox alternative with CRDT state sync, mTLS PKI, and live migration.
Frontier: A Discrete-Event Simulator for Modern LLM Serving
janeyx99 / FlashMLA-ABI-Stable
Forked from deepseek-ai/FlashMLAFlashMLA: Efficient Multi-head Latent Attention Kernels
An AI accelerator implementation with Xilinx FPGA
What DeepGEMM's SwapAB layout is worth on B300, with DeepSeek-V3.2 GEMM shapes in FP8, FP4 and BF16. CUPTI-measured.
DeepSeek Harness: Everything is a Plugin.
Mixture-of-experts (MoE) training megakernel for NVL72s
Repo for ECHO: Efficient KV Cache Offloading with Lossless Prefetching for Serving Native Sparse Attention LLMs (OSDI'26)
Reference implementation and examples of the CuTe Layout representation and algebra.
Autoresearch for GPU kernels. Give it any PyTorch model, go to sleep, wake up to optimized Triton kernels.
MoonEP: A Perfectly Balanced Expert Parallelism Library via Dynamic Redundant Experts
vLLM’s reference system for K8S-native cluster-wide deployment with community-driven performance optimization
We designed a simple harness testing PCB and wrote some firmware for it. Can you figure out everything wrong with it?