Matthew Yue

Software Engineer | Data Scientist | UW Graduate | ML Engineer

Undergrad Research Assistant | University of Waterloo & Purdue University Sep 2023 – Present, Waterloo ON

• Work under Dr. Liu and Dr. Ding on using machine learning to predict carbon emission of cloud platforms (AWS, GCP)

• Built a data ingestion workflow that fetches millions of erroneous or missing ads entities from Retail to Ads domain to ensure data integrity, improve customer experience, and significantly reduce on-call duties.
  • Built AWS CDK to set up workflow infra and configure resources (S3, SQS, CloudWatch) with CloudFormation stacks
  • Implemented workflow in SQL and Scala Spark, developed integration test and manual end-to-end testing with Athena
  • Regionalized workflow to enhance performance, set up alarms to monitor workflow and investigate anomalies promptly
• Migrated launch configurations for product features from error-prone experimental weblabs to launch parameters by marketplace IDs in Java, resolving issues with corrupted weblabs and stabilizing customer portals in the global scope.

• Developed an internal employee evaluation automation system in Python, using Azure, MySQL, and VBA to periodically process reporting emails and aggregate the data for manager’s evaluation, used by hundreds of TD employees.
• Built end-to-end preprocessing and validation pipeline for Azure config onboarding for Database and API ingestion. Used Databricks, PySpark, Pandas, SQL to automate workflow, reduced 2-day work to 1 hour, ensure error-free preprocessing.
• Performed data ingestion and validation on Azure cloud and migrated large datasets to the cloud with Databricks.

• Implemented microservices using React, Django, and Conda to automate email communications, client data collection and processing, converting business workflow to a digital operation model and saving $100K of manual work yearly. 
• Built a relational database for the ML Team using MySQL and Peewee ORM to organize data versions and the ingestion process, enabling faster retrieval of massive image data (several TBs).

• Implemented Modified Bayer Thresholding for image noise removal with Fourier Analysis and a script for parameter tuning using OpenCV, NumPy, Hydra, PyWavelets, and Skimage.
• Developed layer visualization for decoder and auto-encoder in de-aging CNN with Keras and Docker for issue analysis.
• Created an evaluation tool for metrics calculation of neural network models and set up model performance baselines.
• Reimplemented major steps in de-aging network with concurrent multi-processing, increased runtime efficiency by 500%.
• Set up documentation webpage in Sphinx to improve working efficiency and enhance documentation browsing experience.

• Teach first- and second-year students basic programming concept in C and object-orientated programming in C++.

University of Waterloo, Bachelor of Computer Science with Statistics Minor (GPA: 4.0/4.0) Sep 2019 - Dec 2023