Work Portfolio
Jay Tang
MS Business Analytics @ UW Foster ’27 | Former NVIDIA Product Test Engineer | Bridging Engineering Rigor with Data Analytics
About Me
Data-driven Product Test Engineer with 3+ years at NVIDIA, now pursuing a Master in Business Analytics at University of Washington. Improves data review efficiency and test consistency in large-scale semiconductor test data by applying statistical analysis and Python, replacing manual log inspection with self-serve tools used across engineering teams.
Experience
Product Test Engineer
10/2022 — 06/2026NVIDIA — Hsinchu, Taiwan
- Supported 10,000+ wafers per week in mass production by owning and driving wafer sort test programs for data center and GPU devices (TH500, GB206), from development through production release and maintenance.
- Cut average debug time from 1–2 hours to 20–30 minutes by building Python and shell-based tools that enabled engineers to self-serve scalable analysis of large-scale test logs.
- Verified measurement consistency through R², identified root causes using failure analysis (Shmoo plots, DC profiling), and improved test coverage and production stability through Statistical Process Control (SPC) and Pareto analysis.
- Drove wafer production ramp by leading the introduction and bring-up of a new Blackwell GPU testing platform.
- Expanded test capacity by 50%+ through migrating wafer sort test programs to next-generation tester platforms on Advantest SmarTest 8.
Marketing Intern
11/2018 — 06/2019Capsule — Taichung, Taiwan
- Facilitated creator resource exchange and video production via drone and accommodation, and assisted in building a messaging chatbot.
Education
M.S. Business Analytics
06/2026 — 06/2027 (In progress)University of Washington — Seattle, WA
Relevant courses: Data Management and SQL, Probability and Statistics, Data Visualization and Storytelling, Analytics for Firm Decisions.
M.S. Computer Science
09/2020 — 06/2022National Sun Yat-Sen University — Kaohsiung, Taiwan
Space Net Optimization for Single Objective Bound Constrained Problems
08/2025Chun-Wei Tsai, Yi-Cheng Yang, Tzu-Chieh Tang, Che-Wei Hsu
IEEE Transactions on Evolutionary Computation
- Developed the state-of-the-art Space Net Optimization algorithm in C++
- Successfully solved benchmark problems from CEC 2021 and 2022 Single-Objective Optimization Competitions
Space Net Optimization
05/2023Chun-Wei Tsai, Yi-Cheng Yang, Tzu-Chieh Tang, Che-Wei Hsu
arXiv
- Implemented optimization benchmark solutions using C++
- Applied Wilcoxon signed-rank and Friedman tests for statistical evaluation
An Efficient Parallel Machine Learning-Based Blockchain Framework
09/2021Chun-Wei Tsai, Yi-Ping Chen, Tzu-Chieh Tang, Yu-Chen Luo
ICT Express
- Utilized Ethereum blockchain for secure data exchange
- Developed simulated annealing in Python for hyperparameter tuning on MNIST
B.S. Mechanical Engineering (Minor in Computer Science)
09/2016 — 06/2020National Chung Hsing University — Taichung, Taiwan
Relevant courses: Data Structures, Algorithms, Introduction to Data Mining and Machine Learning, Introduction to Information Retrieval.
Project
Food Safety Risk Analysis and Prediction
Python, pandas, scikit-learn, XGBoost, Machine Learning, EDA
- Merged and cleaned two King County open datasets (106K+ inspection records, geocoded via GeoPy) covering 12,261 restaurants, then built exploratory analysis surfacing patterns like repeat-violation rates and risk-category grading gaps
- Engineered a feature pipeline using a chronological train/test split and each restaurant's past violation history only, then trained a shared model across 53 violation types, benchmarking Logistic Regression, Random Forest, and XGBoost, with Random Forest and XGBoost further tuned via validation log loss.
- Built an evaluation framework using Precision@3, Recall@3, F1, ROC-AUC, and Youden's J, outperforming baseline heuristics by ~9.7× vs. random guessing and ~2.5× vs. a naive popularity baseline, converting inspection history into a prioritized checklist for health inspectors.
Large-scale Test-Data Analysis Tool
Python, Streamlit, Pandas, scikit-learn, Statistics, Semiconductor
- Engineers spent hours manually scanning 100,000+ line test logs across tester platforms, wafers, and lots to find the handful of data points that actually mattered for debug and yield decisions
- Built an intuitive, interactive dashboard combining statistical analysis and data visualizations, including R², box plots, Shmoo plots, and bin-distribution/Pareto yield analysis, to parse large-scale test log files and surface root causes for engineering debug and key electrical parameter review
- Self-serve tool replaced manual command-line log inspection, enabling engineers to pinpoint key data points across large-scale data with speed and precision
Personal Interests
