I build and run infrastructure. I turned my lab's mismatched GPU workstations into one SLURM cluster, ran my department's network and its 200 lab PCs for three years, and kept a homelab of my own for years before that. My M.S. thesis put that cluster to work testing self-driving cars in simulation.
I finished my M.S. in Computer Science and Engineering at National Yang Ming Chiao Tung University (NYCU) in August 2026. Through my advisor's collaboration with Taipei Veterans General Hospital, I'm currently building clinical information systems that streamline the care team's workflow. I'm looking for a software engineering role.
M.S., Computer Science and Engineering資訊科學與工程研究所 碩士
National Yang Ming Chiao Tung University · GPA 3.88/4.3陽明交通大學｜GPA 3.88/4.3
Deep Learning A+
Image Processing A+
Computer Vision
UX and Usability Evaluation
深度學習 A+
影像處理 A+
電腦視覺
使用者經驗與易用性評估
2020 – 2024
B.S., Computer Science (honors track)資訊工程學系甲組 學士
National Yang Ming Chiao Tung University · GPA 4.06/4.3 · ranked 7th of 49 · 181 credits陽明交通大學｜GPA 4.06/4.3｜班排 7/49｜181 學分
A+ in 8 of the 9 core required courses, all taught in English; top of the class in Operating Systems and in Computer Organization, where my processor design was the fastest in the class.
Took 11 graduate courses as an undergraduate (33 credits), including:
Network Security
Secure Programming
Advanced UNIX Programming
Intractable Problems
Computer Networks
Video Compression
Numerical Software Development
Perception and Decision Making in Intelligent Systems
Built an offline-capable iPad web app that replaces the hospital's 17-page paper geriatric assessment booklet, designed for Apple Pencil input. The backend is Node and PostgreSQL, deployed with Docker and nginx.
把醫院 17 頁的紙本老年周全性評估表搬上 iPad，做成支援 Apple Pencil 手寫、可離線使用的 web app。後端是 Node 與 PostgreSQL，以 Docker 和 nginx 部署。
TypeScript
Vite
Node.js
PostgreSQL
Docker
2024 – 2026
Graduate student and lab cluster admin研究生、實驗室叢集管理者
Built the lab's SLURM cluster out of mismatched GPU workstations: 11 nodes, 29 GPUs across five models, and 376 CPU cores, with GPU sharding so small jobs can share a card. Storage, accounts, and per-job GPU monitoring are all deployed with Ansible.
Built a companion tool for SpiraPlan, the company-wide project management tool, which was slow and lacked features my team needed, such as bulk edits. I did all of it myself, from design to deployment.
Cached SpiraPlan data so PMs could download it as an Excel file, edit it there, and upload it back; the tool then applied the changes to SpiraPlan in the background.
Ran it with Docker Compose on a VM on the team's ESXi host, with GitHub Actions deploying to production.
Teaching assistant, network and PC teams系計中助教（Net 組、PC 組）
NYCU Computer Science IT Center陽明交大資工系計算機中心
Network team. Ran the department network on Cisco and Juniper switches, kept every switch config in version control, and collected device logs in OpenSearch.
PC team. Managed about 200 lab machines with Ansible and AWX, including playbooks that deploy VM images to Proxmox VE and join them to Active Directory.
Net 組：維運系館網路（Cisco、Juniper 交換器），把所有網通設備的設定檔納入版本控制，並用 OpenSearch 集中收 log。
PC 組：用 Ansible 與 AWX 管理約 200 台電腦教室主機，寫了把 VM image 部署到 Proxmox VE 並加入 AD 的 playbook。
Cisco IOS
Junos
Ansible
Proxmox VE
2022 – 2025
Course teaching assistant課程助教
NYCU陽明交大
Digital Image Processing, Prof. Yi-Ting Chen(fall 2025); wrote the final exam
PCCA winter camp: taught STL and data structures, and set up the online judge
First BPCEI programming camp: taught STL and data structures, and wrote problems and test data
PCCA 冬令營：STL 與資料結構講師，並架設 Online Judge
BPCEI 第一屆資訊營：STL 與資料結構講師，負責出題和生測資
Research研究
A logical scenario: one template with a range for each parameter, sampled into many concrete test runs.邏輯情境：一個模板，每個參數給一段範圍，再取樣成許多具體的測試。
2024 – 2026
PISA: A Distributed Platform for Cross-Simulator Autonomous Vehicle Testing at Scale
M.S. thesis · paper and code to be released upon publication碩士論文｜論文與程式碼將於發表後公開
Runs millions of concrete driving scenarios across different simulators (CARLA, esmini) to test the decision-making of Autoware and other AV stacks. I designed the scaling layer: a Rust task queue, a Go scheduler that submits work to SLURM, and a Python executor that orchestrates simulator and AV containers over gRPC.
Built a 9K-scenario multi-view video dataset by rendering OpenSCENARIO templates from ten viewpoints (ego, drone, infrastructure). Benchmarked vision-language models on classifying the scenarios, with chain-of-thought and in-context prompting.
Industry collaboration with ITRI · not public與工研院合作｜程式碼未公開
Connected ITRI's in-house driving simulator to Autoware so its planning and control could be tested in closed loop. The simulator speaks ROS 1 and Autoware speaks ROS 2, so I wrote a C++ bridge for ego state and detected and predicted objects, packaged as Docker Compose services.
把工研院自研的駕駛模擬器接上 Autoware，讓它的 planning 與 control 能在閉迴路下測試。模擬器走 ROS 1、Autoware 走 ROS 2，所以我用 C++ 寫了 bridge 來傳遞自車狀態與偵測、預測物件，並包成 Docker Compose 服務。
With Prof. Yi-Ting Chen, from my second year大二下起跟陳奕廷教授做
Crashes are rare in real driving data, so we generated accident scenarios from normal traffic plus a target crash type, to test and improve driving models. Worked with the nuScenes dataset and CARLA, and compared against STRIVE, KING, and Social GAN.
真實車禍資料很少，所以我們輸入正常車流和想要的車禍類型，生成對應的車禍場景，用來評測和改進自駕模型。使用 nuScenes 資料集與 CARLA，並以 STRIVE、KING、Social GAN 作為 baseline。
A C++ reimplementation of the nuScenes dataset SDK with Python bindings, for faster indexing and queries on large autonomous-driving datasets.
以 C++ 重寫 nuScenes 資料集 SDK 並提供 Python 介面，加速大型自駕資料集的索引與查詢。
C++
Python bindings
Projects專案
College years大學時期
Homelab個人 Homelab
Started by turning spare PCs into servers in my first year of college. It ran Proxmox VE on ZFS storage behind pfSense and VyOS routers, and hosted my own DNS, mail, VPN, and LDAP. Now retired.
