Guang Ouyang
Studied at University of Connecticut
Works at Axiom
Please note that the recurring weekly time slots shown on my calendar are available on an every-other-week basis. If you're unsure about a specific time, feel free to message me before booking.
Questions? Start chatting with this coach before you get started.
Work Experience
Founder & Principal
Axiom
March 2025 - Present
Built a data-driven decision system that scans hundreds of U.S. equities (scaling toward full market coverage of 10,000+) to identify high-probability pre-breakout setups. - Designed and implemented end-to-end ML/data pipelines from ingestion and validation to modeling, backtesting, and evaluation - Built and maintained a proprietary market and fundamental data store (65+ years, full U.S. coverage) with automated daily (market) and weekly (fundamentals) updates - Developed feature pipelines to detect structured price consolidation patterns - Built predictive modeling framework to estimate expected returns - Constructed backtesting and simulation system with realistic execution constraints (slippage, exit logic) - Optimized system for efficiency under constrained compute environments and parallel scaling Focus: scalable research system for long-horizon, low-frequency investment strategies Machine Learning, SQL and +5 skills
Data Scientist
March 2017 - March 2025
YouTube Ads Revenue Optimization: Led the full-stack ML development cycle in C++ (from feature extraction to deployment) for user language preference models. This initiative resulted in an estimated $400M+ annual revenue gain while maintaining strict Ads targeting latency limits. Core ML Infrastructure: Expanded end-to-end model pipelines into a configurable framework, significantly expediting the development cycle for multiple internal Google projects and improving cross-team velocity. Process Automation: Architected a C++ validation infrastructure to automate weekly user profile refreshes. This system eliminated production errors and saved significant engineering hours by transitioning from manual to automated refreshes. Global Data Strategy: Resolved a major coverage issue in training data by launching personalized language readability surveys in 40+ countries, collecting ground truth data that drove measurable model quality improvements. Ads Attribution & Optimization: Analyzed ads incrementality metrics using multi-touch attribution models (Python/SQL). These insights improved ads conversion rates for partner publishers within the Google Display Ads network. A/B Testing, Exploratory Data Analysis and +6 skills
Data Scientist
1:ID
October 2015 - March 2017
Risk Modeling & Scoring: Developed and deployed production-grade credit and fraud scoring models for leading financial institutions. Leveraged Gradient Boosted Tree (GBT) algorithms to drive predictive accuracy in highly regulated financial environments. Distributed Systems Engineering: Architected a novel distributed clustering algorithm in Spark (Scala) capable of scaling to multi-billion nodes using only 600GB of memory. This innovation enabled high-efficiency user identity resolution at an enterprise scale. Evaluation Frameworks & Monitoring: Established new statistical metrics for evaluating clustering quality in massive-scale networks. Developed an automated outlier detection system to monitor high-throughput data streams for integrity and drift. Machine Learning, Data Engineering and +2 skills
Research Statistician
Pfizer
September 2011 - December 2014
Statistical Data Analysis by SAS/R on real data from Pharmaceutical Industry Research work on applying Proportional Odds Model to ordinal data
Education
University of Connecticut
PhD, Statistics/Data Mining
2010 - 2015
Grade: 4.06/4.3 My PhD research concentration is social network analysis. It covers following topics: 1. Exploring the real power driving social connections Transitivity (friend's friends have higher probability to be friends), and preferential attachment(rich get richer) effects are two commonly know effects driving social connections. I created a novel model providing statistical inference on these effects. 2.Large scale network clustering Developed a novel clustering algorithm that splits large network into multiple closely connected communities. It overcomes the resolution limitation of widely used Modularity-based clustering methods. 3. Influential node detection in social network In plan
Miami Herbert Business School
Master of Science, Mathematics
2008 - 2010
Wuhan University
Bachelor of Science, Mathmatics
2004 - 2008