Systems Engineer ยท Robotics Middleware
Hi there! I'm Yuyuan Yuan.
I build Rust-native middleware for ROS 2 โ working on Eclipse Zenoh and rmw_zenoh, a real-time-capable pub/sub layer that replaces DDS. More on the About and CV pages.
Posts
2026-09-08
Why a Wayland Kiosk Can't Live in a Container
Three fixes, three dead ends, and a root cause underneath all of them: no container runtime can give a process its own virtual terminal.
2026-09-03
Six Days Inside a GPU That Would Not Wake Up
An RTX 5070 Laptop that no NVIDIA driver could initialise, and what it took to prove the bug was in firmware rather than in code anyone can read.
2026-08-21
The Watchdog That Watched a Disk Die Anyway
A disk-space guard observed the exact failure condition it was built to catch, twice, and the disk hit zero nine minutes later. Two bugs, two fixes, three hours apart.
2026-08-01
One Flake, Seven Hosts, and the Backup That Died Without Telling Anyone
Two hosts each carried their own copy of the same backup config, and the same structural mistake in both of them let one backup silently stop running for two days.
2026-07-30
A Nix-Managed GPU VM Is Genuinely Great, and Here's the Proof
Passing an RTX 3060 through to a Proxmox VM, lying to NVIDIA's driver about being virtualized, and making the whole thing reproducible from one command.
2026-07-12
Teaching a Nix-Packaged Photo App About a GPU It Didn't Know It Had
nixpkgs' Immich package has no CUDA wiring. Building it from source cost hours and an OOM kill. Fetching the vendor's own wheel cost minutes.
2026-05-25
Debugging a Bluetooth Kernel Regression on NixOS
How a two-line bounds check broke Bluetooth on the MT7922 chip, and the journey from dmesg to kernel source to patch.
2025-10-21
The NVMe Drive That Only Failed While Charging
Three PCIe kernel parameters fixed nothing. The actual bug was in the firmware blob, and the fix was pinning a five-day-old package.
2017-09-22
Balancing Car
A self-balancing two-wheel robot on a LinkIt 7697 โ state-space control, IMU sensor fusion, and a small reinforcement-learning model.