Skip to content
View thc1006's full-sized avatar
292 followers · 527 following

Achievements

Achievement: Pull Sharkx4Achievement: Pair Extraordinairex4Achievement: StarstruckAchievement: Galaxy BrainAchievement: QuickdrawAchievement: YOLO

Achievements

Achievement: Pull Sharkx4Achievement: Pair Extraordinairex4Achievement: StarstruckAchievement: Galaxy BrainAchievement: QuickdrawAchievement: YOLO

Highlights

Organizations

@kubernetes @zephyrproject-rtos @open-telemetry @nycu-life @nycu-cirda @ARRC-Rocket

Block or report thc1006

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
thc1006/README.md

Hsiu-Chi Tsai · 蔡秀吉

Cloud-Native Telecom · Kubernetes DRA Correctnes

I look for correctness bugs in Kubernetes Dynamic Resource Allocation (DRA) and Cloud-Native Telecom systems. For DRA, that means reading the scheduler and allocator and checking the accounting invariants: whether a device is counted twice, or whether every state mutation made during a failed allocation is rolled back. I also send patches to OpenTelemetry C++, mostly for object-lifetime, cancellation, and concurrency bugs in its exporters and HTTP transport.

When I find one bug, I look for the same bug pattern elsewhere. Many of my patches start that way. I include a regression test with the fix, and reviewers still catch things I missed.

I currently serve as a CNCF Ambassador and as a member of the Nephio Technical Steering Committee (TSC). Nephio is part of Linux Foundation Networking (LFN). I also maintain M5Stack platforms in the Zephyr Project and hold the AMD ROCm Certified Associate credential.

I am an undergraduate at National Yang Ming Chiao Tung University (NYCU) in Taiwan, Studying AI and Medicine and also a little bit Slavic languages. I received clinical training through an Internal medicine clerkship at Taipei City Hospital, Yangming Branch, and Surgical clerkship at University Hospital Hradec Králové.

I grew up in a fishing village in southern Taiwan. I organize the O-RAN in B5G/6G track at COSCUP and give technical talks in Taiwanese. I help and alos want more people contribute to open source so that, together, we can make Taiwan and the World be better and better!!!

Website · Publications · ORCID · LinkedIn · hctsai@linux.com

CNCF DevStats counts 3,943 contributions across CNCF projects. It's update automation and My site lists my open source contributions in detail.

Pinned Loading

  1. ntn-operators ntn-operators Public

    Kubernetes Operators for Non-Terrestrial Networks (NTN) — satellite ephemeris, ground station lifecycle, cell configuration, and terrestrial-satellite failover

    Go 5 1

  2. zephyr-m5stack-sticks3 zephyr-m5stack-sticks3 Public

    First public Zephyr RTOS board port for the M5Stack StickS3 / M5StickS3 K150 (ESP32-S3-PICO-1-N8R8). Board m5stack_sticks3 — runtime-verified boot, console, buttons, BMI270 IMU, and ST7789P3 LCD.

    C 5 1

  3. qwen3.6-speculative-decoding-rtx3090 qwen3.6-speculative-decoding-rtx3090 Public

    First public benchmark of llama.cpp speculative decoding on Qwen3.6-35B-A3B with a single RTX 3090 (post PR #19493 merge, 2026-04-19). 19 configurations covering ngram-cache, ngram-mod, and classic…

    Python 33 2

  4. qwen3.8-speculative-decoding-rtx3090 qwen3.8-speculative-decoding-rtx3090 Public

    Controlled benchmark of llama.cpp speculative decoding for Qwen3.8-27B on one RTX 3090. Built-in MTP at n-max 2 gives +59.8% decode and 37% less energy per request; DFlash2 (PR #27342) reaches +51.…

    Python 1