Angela Fan
Data Analyst
Los Angeles Metropolitan Area United States
Summary
🤩
Rockstar🎓
Top SchoolAngela Fan is a data analyst with 12 years of hands-on experience blending applied machine learning and data analytics across finance, media, and academia, currently based in Los Angeles. She contributes to high-profile open-source ML tooling like Facebook Research’s fairseq—adding multilayer attention mechanisms, LayerDrop pruning features, and encoder fixes—demonstrating deep familiarity with model architecture and training optimizations. A UCLA Cognitive Science graduate specializing in computing and a Master’s student in Data Science at UC San Diego, she bridges human-centered design with scalable data solutions. Her roles range from research assistantships to corporate analytics internships, giving her a track record of translating experimental models into production-ready insights. Notably, she pairs practical data engineering skills with research-minded experimentation, often addressing subtle model and padding issues that improve robustness in sequence models.
12 years of coding experience
2 years of employment as a software developer
University of California, San DiegoUniversity of California, Los Angeles
Github Skills (9)
transformer-models10
pytorch10
machine-learning10
artificial-intelligence10
trainings10
python10
modeling10
model-optimization9
nlp9
Programming languages (5)
Github contributions (5)
facebookresearch/fairseq
May 2018 - Mar 2022
Facebook AI Research Sequence-to-Sequence Toolkit written in Python.
Role in this project:
ML Engineer
ML EngineerContributions:49 commits, 9 PRs, 25 pushes in 3 years 11 months
Contributions summary:Angela contributed to the fairseq toolkit by modifying model parameters for specific models, including writing prompts and stories models. They also added code for a multiscale gated self-attention layer and pretrained fusion models. Furthermore, the user incorporated LayerDrop functionality for training and model pruning, including related code changes and documentation. The user also addressed padding issues within the encoder, demonstrating a focus on model architecture and training optimization.
pytorchnlpsequencepythontransformer-architecture
huihuifan/RBM-CS51
Apr 2015 - May 2015
Contributions:94 commits, 73 pushes, 1 branch in 21 days
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.