CV

Author

Zhe Cui

Published

February 10, 2026

Skills

Category	Details
Programming	Python, SQL, R, Scala, C/C++
Domains	Recommender Systems, Content Quality and Understanding, Multi-sided Marketplace, Ads & Developer Ecosystem
Quantitative	Machine Learning, MLOps, Statistics, Optimization, Large-scale Data Analysis, Causal Inference, Design of Experiments, Large Language Models, Data Modeling/ETL
Technologies	PyTorch, Dataswarm/Airflow, Spark, Presto, Hive, Alerts/Monitoring, Dashboards, R Tidyverse (ggplot2, dplyr, tidymodels), Git/Mercurial, $\LaTeX$

Human Feedback and Preference Systems: Owned the design and rollout of Snapchat’s first large-scale, content-level in-app survey system, defining preference dimensions, ceiling tests, and response models to convert user feedback into durable learning signals.
LLM Evaluation and Standards: Established evaluation methodology and reference tooling for LLM-powered content understanding, including preference-based metrics and model–human disagreement diagnostics used to guide deployment and iteration decisions.
Content Quality and Ranking Systems: Set the technical direction for integrating human feedback, LLM-based understanding, and evaluation signals into short-form video and Discover ranking and policy decisions.
Content Ecosystem Strategy: Defined content quality and ranking strategy at the ecosystem level by translating human preference signals and model evaluation insights into executive decisions on product direction, growth tradeoffs, and platform health.

Improved Advertiser Reliability Metric by 50% with Real-time Machine Learning: End-to-end development, deployment, and monitoring of GBDT models to predict API requests likely to OOM and timeout, dispatching expensive requests to an async tier. Reduced daily advertiser API errors from 6–7% to 3%.
Reduced Ads API Query Latency (p99) by 10%: Applied observational causal inference via propensity score modeling and regression discontinuity to identify features most impactful for slow queries and prioritized fixes with SWE teams.
Predicted Errors Blocking Advertisers: Analyzed and modeled Ads Manager session data to identify error patterns predictive of uncompleted ad creation; findings informed roadmap planning and documentation improvements.
Developed Framework to Estimate Revenue Impact of API Errors: Quantified ad spend loss due to campaign failures to evaluate business impact of SEVs and regressions.
Personalization & Recommender Systems: Led hypothesis-driven analyses and experiments to enhance Facebook App Navigation and Groups Ranking using two-tower neural network models.

Customer Journey Analytics: Designed and implemented an Apache Spark algorithm to model event data for funnel analysis and pattern mining; used derived features to improve churn prediction lift by 20%.
Customer Churn Prediction: Built Random Forest, GBDT, neural network, GLM, and ensemble models to score churn risk and inform targeted retention offers.
Improved Retail Store Internet Speed 10×: Identified outdated connections in demo units, built a business case to upgrade to fiber internet, reducing new customer signup time by 50%.

Resource Allocation in Backhaul-Constrained Small Cell Networks: Developed distributed algorithms in MATLAB and Python for cooperative resource allocation in 5G networks. Published and presented at CISS 2014 (IEE Explore, Paper, Slides).

Apple (2011), Magnum Semiconductor (2011), ON Semiconductor (2008–2010): Projects in signal processing (audio, biomedical), machine learning, and embedded systems.

Masters of Applied Science, Electrical Engineering, University of Toronto (2012–2015) Thesis: Resource Allocation in Backhaul Constrained Small-Cell Networks, Grade A
Bachelor of Applied Science, Electrical Engineering, University of Waterloo (2007–2012), Honors First Class
International Exchange, Electrical Engineering, National University of Singapore (2010)