UC Berkeley CS182 (Deep Learning)
Teaching Assistant (Head of Discussions)     

Led the design of discussion section worksheets in a course with ~250 students. Taught discussion sections of ~30 students twice a week, including topics from ML basics, ConvNet, Graph Neural Net, Transformers to VAE and Diffusion models and more.

 Adobe Inc.
Software Engineer Intern    

Joined FaceLab, ASML team (Applied Science & Machine Learning) and developed Smart Grid for Face Stylization — a project transforming GAN (Generative Adversarial Network) models into a highly-responsive image-editing web app (details under NDA). The underlying tech stack includes TypeScript, LitElement, Webpack for client-side state management, and Node.js, Express, Flask, AWS and PyTorch for the backends.
I also designed a layout algorithm to present 10,000 styles generated from StyleGAN within a 16x16 grid (condensed 97.5% of all styles) based on style encoding affinity scores, by experimenting with UMAP, K-means clustering and linear programming algorithm.

 Berkeley Artificial Intelligence Research
Student Researcher     

Working on an Active Domain Adaptation method (paper) with prof. Trevor Darrell and Devin Guillory, which leverages both data uncertainty and label uncertainty. Implemented an Active Learning framework with modular code for any form of sampling methods. Also working on a View Synthesis project with Diffusion model with prof. Alyosha Efros and Yossi Gandelsman.

 Dell Technologies Inc.
Machine Learning Intern     

Developed 4 machine learning models, such as multi-layer perceptron and Residual Net, to forecast Wi-Fi systems’ throughput. My project attained 88% accuracy in predicting Wi-Fi systems’ performance on 54 Dell laptops and workstations, shortening the development of a Wi-Fi system from 40 to 33 weeks (calculated based on current accuracy) by providing instant forecasts instead of having to wait for prototyping and lab testing. Project developed in PyTorch and Git.

 Vision and Learning Lab @ NTU
Student Researcher    

Conduct a research project on Domain Generalization and Self-supervised Learning (in Computer Vision) at NTU, supervised by professor Frank Yu-Chiang Wang Developed a project on face anti-spoofing (Challenge @ CVPR Conference 2020) with a team of 3. Achieved 97% accuracy (ROC_AUC Score) on the testing set, placed 7th in all 30 submissions to the Kaggle contest (most participants are NTU graduate students).

 Turing Chain Limited
Software Engineer Intern    

 TopScore Education Inc., Taiwan
Full Stack System Developer