Evaluate and Enhance Your LLM Deployments for Real-World Inference Needs

Agent-friendly GPU profile-query CLI

Pie: Programmable LLM Serving

Learn CUDA with PyTorch

A bidirectional pipeline parallelism algorithm for computation-communication overlap in DeepSeek V3/R1 training.

Accepted to MLSys 2026

⭐️ A cross-platform CLI All-in-One assistant tool for Claude Code, Codex & Gemini CLI.

mKernel: fast multi-node, multi-GPU fused kernels

Dynamic Memory Management for Serving LLMs without PagedAttention

An Efficient and Versatile Inference Engine for Distributed LLM Serving

Train the smallest LM you can that fits in 16MB. Best model wins!

LP_Bench

A ChatGPT(GPT-3.5) & GPT-4 Workload Trace to Optimize LLM Serving Systems

MetaAttention: A Unified and Performant Attention Framework Across Hardware Backends(PPoPP'26)

Using a swizzled hierarchical layout for GEMM

Academic Research Skills for Claude Code: research → write → review → revise → finalize

Efficient Long-context Language Model Training by Core Attention Disaggregation

A Jekyll theme for academia

Open-source framework for the research and development of foundation models.

Microsoft Azure Traces

Nex Venus Communication Library

S-LoRA: Serving Thousands of Concurrent LoRA Adapters

Multi-GPU CUDA stress test

Pure Rust + CUDA LLM inference engine

Foundry materializes CUDA graphs along with its execution context to disk to support fast cold start of serving engines.