Undergraduate at PKU, exploring the path from intelligent systems to ML infrastructure.
ML Systems · LLM Inference · Computer Vision · Efficient Computing
- nanoPD — a from-scratch Prefill/Decode disaggregation inference engine for LLMs.
- SLAM3R OnlineCam — real-time dense reconstruction from an online camera stream.
- paperwise — deep reading for research papers with LLM reports, a vector KB, and a knowledge graph.
Collect and merge local Codex usage across macOS, Windows, and WSL. See the multi-device setup guide for first-time setup, SSH authentication, daily updates, and troubleshooting.
Understand the system. Build the system.

