XIAO YOUWEI
Community@uv-xiao
Fourth-year Ph.D. student at Peking University, researching language and compiler techniques for MLSys/Arch/Hardware design.
Agent Skills by XIAO YOUWEI
Showing 54 vetted skills indexed across 1 GitHub repositories.
uv-xiaouv-example-skill
Validate skill mirror and manifest generation in automated tests.
uv-xiaouv-find-skills
Search the open agent skills ecosystem and guide installation via the Skills CLI.
uv-xiaouv-bootstrap-skill-linking
Define prerequisite, companion, follow-up, and escalation relationships between skills.
uv-xiaouv-bootstrap-ml-knowledge-authoring
Scaffold new ML knowledge skills with predefined repository structure and naming conventions.
uv-xiaouv-skill-evolution-manager
Persist session feedback into evolution.json and stitch learnings into SKILL.md.
uv-xiaouv-bootstrap-skill-maintenance
Manage pkbllm skills repository additions, imports, and refactors.
uv-xiaouv-using-pkb
Guide users through discovering and invoking uv-* skills in pkbllm.
uv-xiaouv-start-task
Assemble relevant skill notes into a project's AGENTS.md file.
uv-xiaouv-ml-paper-writing
Creates ML/CS conference submissions with LaTeX templates and citation verification workflows.
uv-xiaouv-serving-llms-vllm
Serve LLMs with vLLM using OpenAI-compatible endpoints and quantization.
uv-xiaouv-tensorrt-llm
Optimize LLM inference with NVIDIA TensorRT-LLM on NVIDIA GPUs.
uv-xiaouv-sglang
Serve LLMs with RadixAttention prefix caching and structured generation.
uv-xiaouv-llama-cpp
Run LLM inference on CPUs, Apple Silicon, and non-NVIDIA GPUs with GGUF quantization.
uv-xiaouv-deepspeed
Guide distributed training with DeepSpeed ZeRO, pipeline parallelism, and mixed precision.
uv-xiaouv-pytorch-fsdp2
Integrate PyTorch FSDP2 fully_shard into training scripts with DCP checkpointing.
uv-xiaouv-ray-train
Orchestrates distributed ML training across multi-node clusters using Ray Train.
uv-xiaouv-slime-rl-training
Run GRPO reinforcement learning training for LLMs with Megatron-LM and SGLang.
uv-xiaouv-miles-rl-training
Train large-scale Mixture-of-Experts models with FP8/INT4 quantization-aware training and speculative RL.
uv-xiaouv-verl-rl-training
Run large-scale RL training for LLMs with verl and GRPO.
uv-xiaouv-moe-training
Train Mixture of Experts models with DeepSpeed and HuggingFace Transformers.
uv-xiaouv-implementing-llms-litgpt
Implement and fine-tune LLM architectures with LitGPT.
uv-xiaouv-mamba-architecture
Implement the Mamba state-space model architecture with O(n) complexity.
uv-xiaouv-rwkv-architecture
Explain the RWKV hybrid RNN-Transformer architecture with linear-time inference.
uv-xiaouv-speculative-decoding
Accelerate LLM inference with speculative decoding and Medusa multiple heads.