CS undergrad @ National Tsing Hua University (Class of 2027)
I optimize scientific and AI applications on GPU clusters, and build LLM-based tooling for EDA. I care about careful benchmarking, correctness, and going deep enough to make things actually work.
📍 Hsinchu, Taiwan · 📫 samuel.zheng0819@gmail.com · LinkedIn
-
ICCAD 2026 CAD Contest — Problem A: LLM-Assisted Netlist Exploration and Transformation (sponsored by Cadence) Building an LLM-based agent that turns natural-language requests into deterministic, verifiable gate-level netlist analysis and transformation operations. Undergraduate research @ CY Lab (PI: Prof. Chun-Yao Wang), NTHU. Repo is private while the contest is ongoing — will be open-sourced afterward. Weekly commits show up in my contribution graph.
-
NTHU Student Cluster Competition Team — HPC application optimization, advised by Prof. Chi-Yuan Chou. Tuning GPU clusters under a fixed power budget and optimizing scientific/AI workloads (ASCOT5, DFTB+). Team code lives in the team organization. Currently coaching junior members for HiPAC 2026; selected to represent NTHU at SC 2026 (Chicago).
| Year | Achievement |
|---|---|
| 2026 | 🏆 Highest LINPACK Award, ISC 2026 Student Cluster Competition (Hamburg) — 335 TFLOPS HPL on 16× NVIDIA H200 NVL under a 6 kW power cap; full marks on HPCG (top score) & the secret application |
| 2025 | 🥇 Champion, APAC HPC-AI Competition — large-scale AI model optimization & benchmarking on HPC clusters |
| 2025 | 🥈 2nd Place, HiPAC — National HPC Application Performance Optimization Competition |
| 2024 | 🥈 Silver Medal, iGEM (MIT-founded, 300+ teams from 40+ countries) — dry lab: PCB design, CAD/3D printing, simulation & mHealth app for project "Pandage" |
| 2023 | 🥇 Gold Medal, 21st Macronix Science Awards (Taiwan's "High School Nobel Prize") — animation-to-comic conversion system, selected from 589 entries |