Skip To Content
Work Experience
Senior Machine Learning Engineer, Content Safety Team at Roblox
Apr 2021 – Current
- Integrate Elasticsearch into a Typescript-written service with around 300k document writes a day.
- Integrate Chinese BERT into TextFilter service.
- Team manager: Kip Kaehler
Full-time Machine Learning Engineer, Content Safety Team at Roblox
Jul 2019 – Apr 2021
- Iterate the latest NLP model, BERT, with increasing labels for real-time low-latency but high-throughput text filtering.
- Build multiple Airflow pipelines to auto-report daily metrics through emails.
- Integrate Hyperscan library in C++ to C# services, which reduces the regex filtering latency by more than 90% (20ms to 2ms).
- Work on novel approaches to detect improper 3D games, including playtime analyses and social network analyses.
- Work on configurable services, based on predefined JSON configs.
- Team manager: Kip Kaehler
Full-time Machine Learning Engineer, Ads Platform Team at Yelp
Mar 2018 – Jul 2019
- Iterated on training pipeline: built features with feedback loops and make generalized model
- Productionized the very first objective model in Yelp and increased lead counts by 111.77% and lower cost-per-lead by 20.98%
- Customized service areas for different businesses and advertisers
- Mentored two industry engineers
- Team managers: Xun, Sundeep
Part-time Machine Learning Engineer, Ads Targeting Team at Yelp
Oct 2017 – Dec 2017
- Built an internal tool using Spark to fetch needed information for analyses on bad advertisements
- Mentor: Eyenaz Alaei
- Team manager: Anusha
Intern Machine Learning Engineer, Ads Targeting Team at Yelp
Jun 2017 – Sep 2017
- Improved model training pipeline
- Add features into Logistic Regression model and reduce cross-entropy
- Mentor: Niloy Gupta
- Team manager: Anusha
Full-time Research Assistant, Academia Sinica
Nov 2015 – Jun 2016
- Finished one paper and later submitted it to PAKDD as first author (refer to: Publications)
- Advisor: Prof. Mi-yen Yeh
Full-time Research Assistant, Intel-NTU Connected Context Computing Center
May 2015 – Nov 2015
- Topic: Intel Smart Car and Driving Behavior Analysis
- Analyzed driving behavior with models written in Python and achieve 0.84 AUC performance
- Presented results with GUI implemented in Python in 2015 Intel Asia Innovation Summit
- Advisor: Prof. Yi-ping Hung
Undergraduate Research Assistant, Machine Discovery and Social Network Mining Lab
Jul 2013 – Present
- Topic: Recommendation of Serial Locations given Check-in Data and Friendship
- Solved one-day route recommendation with improved Bayesian model programmed in Python
- Advisor: Prof. Shou-de Lin
Read more
Education
University of California, San Diego
Sep. 2016 – Dec. 2017
- M.S. in Computer Science and Engineering (CSE)
- GPA: 3.96/4.00 (Operating System: 3.70; Randomized Algorithm: 3.70)
- Selected courses: Latent Variable Model, Deep Learning, Computer Vision, Robotics
National Taiwan University
Sep. 2011 – June 2015
- B.S. in Computer Science and Information Engineering (CSIE)
- Major GPA: 4.04/4.30
- Total GPA: 3.94/4.30
- Selected courses: Software Engineering, Object-oriented Programming, Social Network
Publications
A Classification Model for Diverse and Noisy Labelers, accepted regular paper in PAKDD’17
Apr 2017
- First author paper, cooperated with Kuan Chen
- Derived Graph-based model in C++ and Python to handle annotations from labelers to items
- Advisor: Prof. Mi-yen Yeh and Prof. Shou-de Lin
Other Research Experience
Two-dimensional Proximal Constraints with Group Lasso for Disease Progression Prediction
May 2017
- First author paper, individual work
- Extended multitask learning algorithms from 1D constraints to 2D ones with Matlab and C++
- Advisor: Prof. Mi-yen Yeh and Prof. Shou-de Lin
College Student Research Project, National Science Council
Feb 2014 – Feb 2015
- Topic: Real-time Classification with Missing Data given Model
- Proposed iterative imputing framework in R for data with great amount of missing values
- Advisor: Prof. Shou-de Lin
Knowledge Discovery and Data Mining Cup
Apr 2014 – Aug 2014
- Topic: Predicting Excitement at DonorsChoose.org
- Collaborated with three labs 15 hours a week with models programmed in Python and MySQL
- Ranked 8th out of 472 teams in the most famous machine learning competition
- Advisor: Prof. Shou-de Lin and Prof. Chih-Jen Lin
Read more
Awards and Honors
CFA Exam Level 1 Badge Owner, CFA Institute
Year 2019
Big Data Analytics for Semiconductor Manufacturing, TSMC
Year 2015
- Awards for Excellent Performance out of 124 teams
- Directed a team of three members with R programming language
- Advisor: Prof. Hsuan-tien Lin
ACM ICPC Regional Programming Contest
Year 2013
- Awards for ranked 4th place out of 67 teams
- Solved 8 of 11 challenging coding problems with C++
- Advisor: Prof. Pu-Jen Cheng
Skills
Programming Skills
- Machine learning Libraries
Recent Comments