◦ dbt/analytics/visualizations & reporting development ◦ MLOps: revenue estimation & traffic quality classification◦ Data quality & Airflow management

◦ Facilitated Machine Learning with Large Datasets (graduate/PhD-level, 80+ students) by developing assignments and exams, grading student work, hosting office hours, and directing recitations◦ Key topics: Spark, distributed ML, GPUs, hashing, deep learning optimization, dimensionality reduction

◦ Facilitated Modern Regression (senior/graduate-level, 230+ students) by grading student work, hosting office hours, providing feedback, and proctoring exams◦ Key topics: regression theory, model inference & diagnostics, feature & model selection, regularization

◦ Engineered 120+ features from sales and CRM data (19+ million rows, ~10.5 GB) with aggregation and NLP methods to determine optimal client contact strategies to increase mutual fund inflows◦ Built and automated data pipelines with PySpark on Databricks to streamline ML development, established benchmark model performance of 93.7% accuracy predicting purchase activity & suggesting next steps◦ Segmented client base based on purchase history with regression and determined best combinations of client interactions across client types for maximizing conversion rates◦ Presented key findings to management, upgrading sales’ targeted outreach & client retention capabilities

◦ Built frameworks for evaluating craving-predicting ML models through F-score and K-fold cross-validation, facilitating context-specific model evaluation◦ Parallelized hyperparameter tuning process of CNN and XGBoost models by incorporating multiprocessing, decreased evaluation runtime by up to 800%◦ Employed and tested multiple feature engineering techniques (Gaussian Smoothing, Autoencoding, etc.) to enhance model performance, increased cross-validation accuracy by 3.5%◦ Created 10+ model performance visualizations (feature importance, validation accuracy over time, etc.) that were integrated into investor relations presentations, improving brand image and company outlook

◦ Developed data pipelines involving NLP, mixed-effects modeling, and hypothesis testing that determined the association between course material presentation methods and student performance◦ Led the data analysis team of Podsie, an educational start-up, and analyzed the efficacy of its products under low-sample size constraints and presented findings to board of directors◦ Formatted and designed 15+ reports of critical findings and visualizations to research leads and grant reviewers, leading to continued funding for research projects at the organization