Seattle, Washington, United States
I’m an Applied AI Data Scientist with nearly a decade of experience building and deploying machine learning solutions that drive measurable business impact and product growth. I excel at the intersection of AI innovation and practical implementation, having led projects spanning causal inference, predictive modeling, generative AI, and computer vision across e-commerce, real estate, edtech, and CPG verticals. As both a hands-on practitioner and strategic thinker, I thrive in the player/coach role delivering sophisticated technical solutions while guiding teams toward impactful outcomes. My approach combines rigorous analytical methods with business intuition, allowing me to translate complex data into actionable intelligence for stakeholders across organizational levels. Whether fine-tuning large language models, developing causal inference frameworks, or creating computer vision solutions, I bring both technical depth and cross-functional collaboration skills to solve challenging problems. I view my role as similar to a point guard in basketball: surveying millions of data points in real-time, identifying opportunities for success, and setting up actions that put teams ahead of the game. Consider me a data alley-oop specialist ready to elevate your organization's AI capabilities.
I lead applied ML and causal inference work for TikTok Shop US, focusing on problems like predicting customer lifetime value, measuring what actually moves the needle on campaign performance, and helping creators find products their audiences will genuinely want (rather than just chasing commission rates). Lately I've been deep in AI agent development, exploring what it looks like to have an AI co-pilot that can actually do data science alongside you.
- Helped a global beverages company identify underperforming stores by developing, testing and deploying a generalized boosted regression model to find opportunities in certain spirit categories across thousands of stores utilizing location-based customer data, consumer demographics and store demand - Determined a methodology for ensuring better distributor compliance for a global beverages company with shelf placement for different liquor brands by building a convolutional neural network using Tensorflow to correctly identify bottles with over 95% accuracy from photos taken by sales team members across thousands of locations nationwide - Optimized in-store brand channel spend for a global beverages company by identifying the effectiveness of in-store visits by field and distributor representatives while also calculating potential store lift for targeted activities ranging from sampling events to floor displays - Improved sales forecast accuracy by 15% for an international nutraceutical company by analyzing hundreds of thousands of sales with ensemble machine learning algorithms in Python - Deployed a machine learning model for an international nutraceutical company using the scikit-learn library hosted on Amazon Web Services increasing inventory turnover by 10% through optimization of sales projections - Utilized k-means clustering to identify similar customers across a varied product portfolio helping a beverage company optimize their social media advertising to increases their conversion rate by 40% - Designed and developed a Swift-based iOS app utilizing Firebase as a BaaS for users to track wellness goals and earn points that were redeemable for rewards from local businesses - Managed the design and development of a Unity 2D-based mobile app utilizing Amazon Web Services to enable users to track working memory performance while using a natural supplement by playing a polygon-matching game
- Pioneered a novel approach to evaluate Grammarly's referral program impact, overcoming traditional A/B testing limitations by integrating Propensity Score Matching (PSM) with an Artificial Counterfactual Estimation (ACE) model. This method involved matching users based on observable characteristics and engagement behaviors, then estimating the program's effect with a machine learning-based causal impact analysis. This strategy enabled a nuanced understanding of the program's effectiveness, tailored to diverse user segments and behaviors, significantly enhancing targeting and optimization efforts. - Conducted a comprehensive causal inference analysis to measure the incremental impact of launching GrammarlyGO, an AI writing assistant. Utilized Bayesian time series models to predict counterfactual scenarios and employed difference-in-differences estimation, coupled with Dynamic Time Warping, for precise control and treatment group comparisons. - Designed a predictive machine learning model to forecast 4-week user retention based on initial user actions, employing XGBoost and Shapley values to identify key factors influencing retention. Implemented strategic product interventions that targeted these factors, significantly boosting user engagement and retention rates. - Engineered a subscription churn prediction model, leveraging a comprehensive dataset encompassing product engagement and user demographics to accurately forecast churn likelihood two months prior to renewal dates. This significantly reduced churn rates and enhancedcustomer lifetime value. - Led the fine-tuning of Llama LLM architecture for Grammarly's AI Writing Tools suite, optimizing model performance for the Cover Letter Generator, Executive Summary, Business Report, and Job Description tools, resulting in higher-quality content generation that maintained consistent tone and contextual relevance across diverse professional writing formats.
- Led the development of a deep learning architecture to accurately detect camera height and adjustments during panorama capture for 3D home tours, resulting in a reduction in panorama rejections and enhancing the quality of virtual property showcases. - Developed an object recognition model leveraging listing descriptions to extract and highlight key features (e.g., marble countertops) in 3D home tours, using NLP techniques and machine learning frameworks. This approach enhanced virtual tour engagement by spotlighting property highlights effectively.
- Designed and deployed a machine learning model improving pricing accuracy for analysts on "For Sale" properties. - Designed a Keras-based multi-input/multi-output model for weighing average valuations of similar properties for analysts’ final valuation. - Researched and built a DistilBERT model for feature extraction and validation from listing descriptions, Multiple Listing Service records and county property data to better identify property amenities such as pools and fenced yards.
- Built a multi-touch attribution model for marketing stakeholders using Markov chains to consider channel sequence while assigning transition probabilities to each conversion state based on the likelihood of successfully converting customers to sales - Designed and conducted multiple incrementality tests using Bayesian structural time-series models to estimate the causal effect of designed interventions on test markets while utilizing dynamic time warping to do time series matching with set criteria for determining control market matches - Conducted feature engineering and created a machine learning model, utilizing XGBoost, to predict the likelihood of a renter to default on their rent payment throughout the course of their lease
- Created metrics for measuring marketing communications program effectiveness through lead generation/sales attribution, customer brand awareness and potential employee brand awareness - Administered the entire marketing communications department budget, determining resource allocation for activities across the business based on internal ROI assessments - Supervised all of GE Transportation's digital activities including getransportation.com, the company's online advertising efforts and the company's social media channels as well as supporting Transportation digital initiatives conducted through corporate digital channels - Led the development efforts of five individuals for four different enterprise sales and event applications in Android and iOS that reached an audience of over 50,000 individuals while incorporating multiple external APIs
- Managed the redesign and redevelopment of getransportation.com in Drupal, which included reorganizing content, building new functionality and conducting user research with over 500 internal and external stakeholders leading to an increase in monthly site traffic by 25,000 users - Created metrics for measuring marketing communications program effectiveness through lead generation/sales attribution, customer brand awareness and potential employee brand awareness - Directed the development of all marketing collateral to ensure a unified branded user experience for customers across a number of different industries