Emma Zhou

Analytical thinker with interdisciplinary expertise in data science, behavioral economics, and product innovation. Proven ability to bridge quantitative modeling and capital narratives across B2B and AI-native products.

• Financial Psychology
• Narrative-Driven Investment
• Cross-Cultural Systems

Impact of Educational Investment on GDP in East Asia

Self-Initiated Research | Tools: Python, World Bank API, Data Visualization

· Collected and analyzed macroeconomic data from World Bank and national sources to evaluate whether public education spending drives GDP growth across East Asian countries.

· Conducted cross-sectional regression and visual analysis; findings suggest limited direct correlation between education investment and short-term GDP growth in the region.

· Developed transferable skills in real-world data extraction, policy impact evaluation, and hypothesis testing.

Product Analysis of Hume AI’s Emotional Intelligence Capabilities

Self-Initiated Research | Tools: Competitive Research, PRD Writing, UX Mapping

· Analyzed Hume AI’s emotion-sensing API products and constructed a structured PRD to evaluate functionality and business use cases.

· Benchmarked competitors (e.g. Affectiva, Replika) to understand strategic differentiation.

· Informed downstream product strategy and AI-emotion integration in user-facing experiences.

Predictive Modeling of Electoral Outcomes in U.S. Primaries

Data 102 Final Project, UC Berkeley | Tools: Python, Scikit-learn, Causal Inference

· Investigated two core questions: (1) Do endorsements increase vote share? (2) Can demographics predict election wins?

· Built causal models (ATE via linear regression) showing that endorsements increase vote share by ~15.4%.

· Trained logistic regression and random forest classifiers on candidate demographics (race, gender, incumbency, Trump/Sanders endorsements); achieved 81–82% accuracy with strongest predictors being incumbency and Trump endorsement.

· Enhanced skills in feature engineering, model evaluation (AUC, F1-score), and bias-aware political data modeling.