As the most senior data scientist at the company, I designed and implemented machine learning and experimentation products for restaurants and brought leadership to the data science team.- Deployed machine learning models to production. (Python, Databricks)- Built models to identify restaurant customers likely to churn and their predicted spend, allowing client companies to better target marketing efforts. (Python) - Devised configurable recommendation systems to drive personalized marketing campaigns for Brightloom clients. (Python)- Formulated experimental designs and determined appropriate statistical methods to measure incremental value for Brightloom clients. - Built an A/B testing platform to run experimentation at scale in production. (R, Databricks) - Collaborated with data platforms engineering to determine the appropriate tech stack for the data science team and prototype model deployment options. (S3, EC2, Databricks, Python, R)- Interacted directly with Brightloom clients to better determine their needs and potential data science solutions.

Built machine learning solutions to improve supply chain performance & provide personalized shopping experiences to each customer. - Collaborate with stakeholders to solve business problems using machine learning and evaluated the impact on relevant business KPIs. - Led the predictive modeling process end-to-end: explored and cleaned transaction and fulfillment data of 100M+ rows from a SQL database, engineered model features, carried out feature selection, chose algorithms, and tuned hyperparameters. - Built a GBM quantile regression model to predict customer order shipping dates, which improved accuracy by 20%. - Increased accuracy of product level demand predictions by 46% (~3K new products offered daily) by building a GBM model. - Collaborate with engineers to improve Zulily's personalization services by adding new features to our models, exploring algorithm options, and employing Bayesian optimization for hyperparameter tuning.- Mentor teammates in all steps of the modeling process.Designed and implemented A/B tests.- Collaborate with stakeholders across the business to determine best experimental designs and metrics to measure. - Improve the internal A/B testing platform by decreasing system runtime by 80%, stabilizing the system by introducing package management, and implementing new features to improve information availability for stakeholders.

- Deployed a Chrome extension, GameOn, that shows consumers the predicted complexity of a boardgame when browsing games online.- Retrieved data on 220,000+ board games through the boardgamegeek API.- Employed feature engineering and hyperparameter tuning to build a random forest classifier of game complexity with 72% accuracy.

- Developed statistical approaches to improve fisheries resource management.- Constructed a Bayesian hierarchical model that integrated 4 data sources to predict the effects of environmental change on Pacific hake and Pacific herring.- Developed a new random effects mixture model that improved accuracy and precision of resource usage by up to 75%.- Facilitated a team of 6 people to analyze collected measurements on fish and zooplankton.- Trained and managed 3 technicians.- Received a Dept. of Defense Science and Engineering Fellowship to pursue work on improving statistical methodologies.- Planned, organized, and carried out week long fieldwork trips on a small fishing vessel.- Lived and worked in tight quarters with field team members to carry out field work operations.

-Teaching assistant for an ecological modeling course. -Led lab and skills sections to teach students basic modeling concepts.-Worked with students one-on-one to help them understand concepts they were struggling with.

- Built an individual-based model of migration routes of Atlantic salmon, an economically important species.- Predicted salmon locations in the Gulf of Maine to inform management efforts.