At the 2024 Google Taiwan Hardware Product Sprint, I led the software design and development of PiBuddy, a multifunctional companion robot, which won the Best Engineered Product award. PiBuddy is equipped with features including a camera, voice recognition, a to-do list manager, and interactive emotional responses. Over the three-month sprint, I:• Designed the overall software architecture, ensuring smooth concurrent operation of PiBuddy’s hardware components through multithreading and multiprocessing.• Developed 11 page classes and handlers for key functions, including the emotion engine, widgets, battery, and encoders, optimizing PiBuddy’s user interface and emotional interactions.• Created a modular handler system for efficient communication between hardware groups, streamlining robot control and expanding future scalability.• Led the development of a custom SQL database for managing user data like to-do lists, photos, and videos with a robust reset option for easy system recovery.• Designed PiBuddy’s emotional feedback system, integrating hardware sensors and voice inputs to create an engaging and interactive user experience.The project involved extensive hardware-software integration and collaboration, resulting in a robot that serves as both a productivity assistant and a comforting companion for users.

During my internship at Intel, I spearheaded the development of a Python-based system from the ground up, designed to monitor foundry, test, and performance data for Intel’s key products, including Alder Lake, Meteor Lake, and Arrow Lake. This system evolved into a fully operational tool, leading to the hiring of two additional engineers to support its expansion. The project significantly reduced data processing time, saving 5 hours per week for a group of 30+ engineers, thereby enhancing team productivity.• Built a Python application to monitor product performance, enabling more efficient decision-making.• Optimized SQL queries to improve data retrieval times from large datasets• Supported the data pipeline for automated ingestion and transformation, ensuring smooth data flow.• Cleaned and organized raw data, improving data quality and usability.• Collaborated with cross-functional teams to gather requirements and deliver actionable insights.• Delivered two workshops to 100+ engineers, sharing best practices and fostering knowledge sharing.

• Developed bipartite matching-based method achieving a 96% identification accuracy (less than 1% drop) for up to 50 single-typed faults in 3 and 4-layered MNIST models• Conducted data analysis on over 5000 simulated chip failures, optimizing fault injection algorithm.• Devised a training workflow increasing the accuracy of faulty models by up to 40% for MNIST recognition.• Composed research papers and submitted research results to the IEEE 32nd Asian Test Symposium as first author.

• Extracted data of 27 corners and 270 combinational/sequential cells from .lib files to .csv files for feature characterization.• Developed and refined machine learning models to predict cell timing behaviors, achieving a mean of 93.5% accuracy on test sets.

• Contributed to the publication pool of National Taiwan University by developing personalized feedback strategies and guiding12 students through academic writing process.• Identified and rectified six key issues in the application website, streamlining the application process.

• Developed mathematical models to optimize the utility function of Non-Terrestrial Networks (NTN) providers using game theory contract theorem.• Created and implemented solution algorithms in polynomial time complexity through Python, Numpy, and Matplotlib.

• Led the design and construction of the company's official website.• Provided UI/UX input for major product in 36.8% of GitHub contributions, contributing to increases in exposure and click rate of 636% and 558% respectively.