Samuel Kwong San Francisco Bay Area

Email: samjkwong@gmail.com linkedin.com/in/samjkwong

samjkwong.com

Education
Stanford University Stanford, CA
•
Master of Science – MS, Computer Science Apr. 2019 - Dec. 2021 (Expected)
Stanford University Stanford, CA
•
Bachelor of Science – BS, Computer Science Sep. 2016 - Apr. 2020 (Graduated)

Experience
Google Sunnyvale, CA
•
Software Engineer Intern Jun. 2020 - Sep. 2020
◦ Team: Google Cloud TPU AutoML
◦ Project: Implemented deep learning models, ResNeSt and RegNet, in TensorFlow (originally in PyTorch and MXNet) to run
novel experiments measuring their performances on Google TPUs for direct comparisons to Google’s EfficientNet-X. Attained
74.7% throughput increase with my ResNeSt implementation migrating from NVIDIA Tesla V100 GPU (float32) to TPU v3
(float32) and 120% increase migrating to TPU v3 (bfloat16).
◦ Using: Python, TensorFlow, Keras, GCP
Verkada San Mateo, CA
•
Software Engineer Intern Mar. 2020 - Jun. 2020
◦ Team: People/Vehicle Search
◦ Project: Tasked with improving facial recognition and vehicle detection for Verkada’s enterprise video security camera
system. Worked on AWS backend and computer vision pipeline. Deployed to production feature that prevents occluded faces
from being matched to other occluded faces as false positives – particularly faces with masks since the pandemic outbreak.
◦ Using: Python, AWS, Docker
Stanford Artificial Intelligence Lab Stanford, CA
•
Graduate Student Researcher Sep. 2019 - Present
◦ Group: Stanford Vision and Learning Lab – Video Understanding (Prof. Fei-Fei Li, Dr. Juan Carlos Niebles)
◦ Project: Ongoing computer vision research in video action recognition and dataset scene-bias mitigation. Implemented novel
model to avoid scene bias when making human action classifications from video input.
◦ Using: Python, PyTorch
Stanford University Stanford, CA
•
Research Intern Jun. 2018 - Aug. 2019
◦ Workplace Health Innovation Lab: Developed computer vision camera software to automate real-time human ergonomic
evaluation. Built Python pipeline for 3D pose estimation and real-time analytics to improve subject’s ergonomic posture.
◦ Bill Lane Center for the American West: Developed NLP models to measure legislative patterns for the CA coastline.
Developed Python web-scraping applications to track progress of CA Prop. 50 Funding for Public Water Systems.

Teaching Assistantships
CS 271: Artificial Intelligence in Healthcare Stanford University
•
Graduate Teaching Assistant Autumn 2020-2021
◦ Responsibilities: Design/teach graduate-level course for cutting-edge deep learning approaches in healthcare tasks using
image, text, multimodal and time-series medical data; instructed by Prof. Serena Yeung.
CS 41: The Python Programming Language Stanford University
•
Teaching Assistant Winter 2019-2020
◦ Responsibilities: Taught graduate/undergraduate students fundamentals of Python and data-analysis libraries for ML.

Projects
• Extreme Low-Resolution Action Recognition with Cross-Resolution Knowledge Distillation: Deep learning project to
address analyzing human behavior while hiding their identity by lowering the video input resolution. Experimented with novel
approach using cross-resolution knowledge distillation between high-resolution and low-resolution videos.
• MapReduce Implementation: Implemented a MapReduce framework, harnessing multiprocessing, networking, threads,
concurrency, distributed computing, a central CPU cluster and a shared file system.

Skills
• Languages: Python, C++, C, Java, JavaScript, SQL
• Technologies: TensorFlow, Keras, PyTorch, Linux, GCP, AWS, Docker, CircleCI, Sentry, OpenCV, React, Git
• Research Experience: Computer Vision, Natural Language Processing, Deep Learning/AI/ML
(Last updated Sep. 2020)