ServiceNow
Senior Machine Learning Developer, Toronto (October 2021 ~ Present)
Focused on building and maintaining pipeline of schedule optimization product for ServiceNow platform.
-
Software Engineering: Implemented and tested the pipeline of schedule optimization product.
-
Software Engineering: Refactored pipeline code from legacy code base to make it easier to understand and maintain for long-term.
-
Scalable Data Pipeline: Experimented and implemented muti-process to handle optimization
scheduling jobs that need to be run within certain time-limit.
-
Software Design: Actively participated software designing and innovation workshops to
review design proposals.
-
Documentation: Wrote formal documentation for code design and process by Sphinx to
internal and external users.
TradeRev
Machine Learning Engineer, Toronto (March 2019 ~ October 2021)
Focused on vehicle pricing system and recommendation system with production-level
implementation on AWS.
-
Machine Learning: Refined and improved the performance of vehicle pricing
system (with more than 900 serving machine learning models) and recommendation
system on AWS.
-
Machine Learning: Built an evaluation platform for internal data scientists to evaluate their experiments on machine learning projects. The metrical results/charts/graphs are displayed on Periscope dashboard for quick review. Monitored the performances of all machine learning models in production on DOMO dashboard.
-
Software Engineering: Implemented and maintained source and infrastructure
code of vehicle pricing system and recommendation system.
-
Data Engineering: Wrote Snowflake SQL queries to generate reports from Snowflake platform for business and marketing departments to make decisions.
-
Research & Algorithms: Tested hypothesis for loss risks in vehicle pricing
system and participated in the process of designing better algorithms/models to
minimize monetary loss risks.
-
Testing & Documentation: Designed unit/integration tests to validate pipeline
code of vehicle pricing system and wrote detailed documentations on Confluence.
-
Demo Picture
Ecobee
Data Science Co-op, Toronto (May ~ December 2018)
Built a data pipeline and deep learning models for sound classifications on Google Cloud
Platform.
-
Machine (Deep) Learning: Built and tested different kinds of machine
learning models (decision tree, random forest, XGBoost, etc.) and deep learning
models (CNN, CRNN, etc.) to test the performances of indoor (people talking, human
movements, household activities, office environment, etc.) and outdoor (car-horn,
siren, construction, and vehicle passing) sound classifications. Designed a sound
classification model (CNN + random forest) using Python TensorFlow and Keras to
classify indoor and outdoor sounds with about 88% overall accuracy within respective
experiment tests.
-
Data Engineering: Collected related sound data by designing experiments in
the Ecobee sound lab; acquired most public open-source sound data (DCASE, ESC-50,
etc.) and Google AudioSet (about 2 million sound samples and 620 classes);
implemented data augmentation to increase the variety of input sound data.
-
Data Pipeline: Implemented a scalable inference pipeline with Apache Beam on
Google Cloud Platform (GCP) framework to parallelly process sound input, send prediction
requests to ML Engine on GCP, and further handle prediction results for users in real-time.
-
Software Engineering: Used TensorFlow and Keras to build the deep learning
models and deployed machine learning models on ML Engine on GCP to fulfill
prediction requests.
-
Testing & Documentation: Tested the built data pipeline and sound
classification models with evaluation metrics such as accuracy, precision, recall,
F1 score, etc. Maintained Jira tasks’ tracking, confluence reports, and the GitHub
repository for the sound classification project in weekly basis.
-
Demo Link
PRACSYS LAB, The Computer Science Department, Rutgers University
Summer Software Developer, New Jersey, U.S.A. (June ~ August 2016)
Built a visualization of large crowds in transportation hubs.
-
Software Engineering: Utilized the 3D-Graphic Unity Game Engine and C# to
develop the visualization software, which can simulate virtual moving agents and
buildings in real-time.
-
Research & Algorithms: Solved the large real-time data handling problem by an
innovative multithreading-design to simultaneously read and update data for as many
as 6,000 virtual agents with animations.
-
Research & Algorithms: Designed a dynamic memory management method to handle
large input data (about 20 GB per 1 hour) in the simulation for long-time. As a
result, the real-time simulation can keep running for hours without running out of
memory.
-
Demo Link
University of Toronto, St. George Campus, Canada
Master of Science in Applied Computing, Cumulative GPA: 3.68/4.0 (January 2019)
-
Courses: Introduction to Machine Learning & Data Mining, Foundations of Computer
Vision, Computer Graphics, Object Modeling & Recognition: Shape Perception in Human
& Computer Vision
-
Teaching Assistant for CSC 209 Software Tools and System Programming (Fall 2017 &
Winter 2018)
Rutgers University, New Brunswick, New Jersey
Bachelor of Science in Computer Science with Mathematics Minor (May 2017)
-
Cumulative GPA: 4.0/4.0, Highest Honors, Summa Cum Laude
-
SAS Excellence Awards with the Clark H. Johnson Endowed Scholarship (2015-2016 &
2016-2017)
University at Buffalo, SUNY, New York
Master of Science in Economics (May 2014)
The University of Nottingham, Ningbo China
Bachelor of Arts in Finance, Accounting and Management (July 2011)