Cathy Deng
Staff Engineer, AI Residency at Propel, Inc
San Francisco, California, United States
Summary
🤩
Rockstar🎓
Top SchoolCathy Deng is a Staff Engineer in AI Residency based in San Francisco with 12 years of experience building production-grade systems and machine learning tooling. She has a strong background in data quality and entity resolution, contributing core features and tests to the widely used dedupe Python library and implementing canonicalization algorithms for record linkage. Her career spans growth and product engineering roles at companies like Patreon, Promise, and Incredible Health, plus hands-on data-science work at DataMade where she built ML-based address parsers using CRFs and OpenStreetMap data. Cathy pairs mathematical training (BA, cum laude, Northwestern) with pragmatic engineering, frequently moving between backend systems, data cleaning, and feature engineering to deliver reliable ML pipelines. An engineer who bridges research and production, she quietly improves downstream data quality with thoughtful algorithms that reduce duplicate and noisy records at scale.
12 years of coding experience
11 years of employment as a software developer
B.A. (Cum Laude) Mathematics, B.A. (Cum Laude) Mathematics at Northwestern University
Github Skills (20)
strings10
python10
data-science10
address-parser10
text-manipulation10
feature-engineering10
numpy10
data-preprocessing10
dataprep10
parse10
clustering10
data-structure9
openstreetmap9
algorithm9
algorithms9
Programming languages (11)
Github contributions (5)
datamade/usaddress
Jul 2014 - Sep 2016
Role in this project:
Data Scientist
Data ScientistContributions:197 commits, 6 PRs, 50 pushes in 2 years 2 months
Contributions summary:Cathy primarily contributed to building a machine-learning-based address parser. They implemented data preparation scripts for training and testing, utilizing the pycrfsuite library. The user also developed a script to predict address tags and experimented with feature engineering, including the addition of previous and next word features. Furthermore, they integrated OpenStreetMap (OSM) data to create training files.
python-librarynlppythonstringsaddress
dedupeio/dedupe-examples
Jul 2014 - Dec 2015
:id: Examples for using the dedupe library
Role in this project:Back-end Developer & Data Scientist
Back-end Developer & Data ScientistContributions:13 commits, 2 pushes, 1 comment in 1 year 5 months
Contributions summary:Cathy focused on implementing and refining a canonicalization feature within the `dedupe-examples` repository. Their work involved developing a `getCentroid` function to determine the best representation of an attribute, followed by a `getCanonicalRep` function. They utilized the `dedupe` library for entity resolution and record linkage, including string distance calculations, demonstrating a strong understanding of the dedupe library's functionalities, and relevant skills in Data Science and data cleaning.
record-linkagepythondedupededuplicationentity-resolution
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.