Zhijie Xia

Blog

Why I pre-ordered a Framework Laptop 13 Pro

Leaving Huawei, losing access to H200s, and the long search for a laptop I would never have to replace.

3 min read
hardwareframeworklaptopgpu
Enabling MXFP8 Inference on Ascend NPUs for VERL

How I brought MXFP8 quantization to VERL rollout on Huawei Ascend 950 devices — two PRs, a weight reloading puzzle, and a lot of tensor shape debugging.

4 min read
verlvllmascendquantizationmxfp8rlnpu
Training-Inference Discrepancy in LLM Reinforcement Learning

How quantization in inference engines secretly turns on-policy RL into off-policy training, and how adaptive control can fix it.

3 min read
reinforcement-learningllmrlquantizationtraining
Cuda implementation for mutiplication to illustrate CUTLASS

A deep dive into CUDA matrix multiplication and how CUTLASS optimizes GPU kernels.

3 min read
cudacutlassgpuhpc
CPSC 599 Competitive Programming

Course notes and problem-solving strategies from CPSC 599 Competitive Programming at the University of Calgary.

1 min read
competitive-programmingcourse-notesalgorithms
September 2023 Daily Leetcode: Grind 75 in Python

Working through the Grind 75 LeetCode problems in Python during September 2023.

13 min read
leetcodepythongrind-75algorithms
Crash course on tmux

A quick reference guide and crash course on using tmux for terminal multiplexing.

1 min read
devopsterminaltmuxproductivity
Feb 2023 Daily Leetcode: Leetcode in C++/RUST

Daily LeetCode problems solved in C++ and Rust throughout February 2023.

15 min read
leetcodec++rustalgorithms
Advent of Code 2022

My solutions and walkthroughs for Advent of Code 2022 challenges.

1 min read
competitive-programmingadvent-of-codealgorithms
Jan 2023 Daily Leetcode: Leetcode in Rust/C++

Daily LeetCode problems solved in Rust and C++ throughout January 2023.

40 min read
leetcoderustc++algorithms
Hello World

Welcome to my personal blog — a place for notes on reinforcement learning, LLM infrastructure, and whatever I am building.

1 min read
blog