I am currently an SDE at Meta, based in the bay area. I received my Master’s degree from CMU’s Master of Computational Data Science (MCDS) program. I graduated from Columbia University in 2017 with a double major in Computer Science (AI track) and Statistics. In the summer of 2022, I interned at Meta's Ads Responsibility and Privacy team working on hierarchical text classification. Previously, I worked as a business analyst intern at Amazon and a data scientist intern at UNIQLO.

I enjoy working at the intersection of ML and software engineering, with prior experience in developing and deploying large-scale ML models. My current areas of interest are large language models (LLM), generative AI, multimodal ML, and statistical ML. I am proficient with various programming languages (Python, SQL, C++, Java, R), deep learning frameworks (PyTorch, TensorFlow), data analysis libraries (Pandas, Numpy, matplotlib), and cloud computing platforms (AWS, Azure, GCP). I also have experience with handling large-scale data with Spark & Hadoop and developing containerized applications using Docker and Kubernetes.

🏠Return to homepage

CONTENTS

👩‍💼 Work Experience

Software Engineer, Meta

Feb 2023 — present

Software Engineering Intern, Meta

May 2022 — Aug 2022

Multilabel Text Classification

• Trained an XLM-R multilabel text classification model that improved AUC-ROC for 63% of classes compared to the production model;
• Engineered a daily data pipeline and deployed model to offline serving

Hierarchical Classification

• Developed a hierarchical classification model with per parent node local classifiers which boosted performance for 80% of the minority classes;
• Reduced computation cost and inference latency using a shared encoder architecture
• Designed an ML experiment framework to automatically launch and manage model training and inference

Business Analyst Intern, Amazon

April 2021 — Aug 2021

• Automated Australia manual transfer with Python and Bash to save 10 hours of labor per week
• Extracted, Transformed, and Loaded (ETL) shipment data (40M rows) into AWS Datanet & Redshift with MySQL
• Conducted root-cause analysis of data quality issues

Data Scientist Intern, UNIQLO

June 2020 – March 2021

• Improved CTR by 200% in production, achieved 60.6% recall at top 30 with product recommender system trained on 24M order data; Implemented offline serving pipeline using multi-threading
• Developed a stockout forecast model using GBDT with Python, achieved 90.9% AUC - ROC
• Implemented web scraping and RPA with requests, selenium, and BeautifulSoup
• Performed ETL with Hive and designed visualizations with Tableau to address business questions

👩‍💻 Projects

Intelligent Music Generation

https://github.com/ZhouyaoXie/Intelligent-Multimedia-Art-Generation

Jan 2022 – Apr 2023

Led a 4-people team to develop MusicCLIP, a Transformer-based text-conditioned music generation model.

The model has an encoder-decoder architecture, with the following components: a MuseMorphose music encoder, BERT text encoder, LXMERT cross-attention blocks, MuseMorphose music decoder. The model is trained on contrastive loss.

Twitter Big Data Analytics Web Service

Nov 2021 – Dec 2021

• Performed ETL on a Twitter dataset (>1TB) with Spark into MySQL and loaded database into AWS RDS
• Collaborated within a group of three to develop and containerize a user recommendation web service with Java and Kubernetes