Seattle, Washington, United States
I am a Decision Scientist who has worked within Operations Intelligence and Staffing Forecasting. I specialize in simplifying complex business problems for the Operations organization, and transforming them into clear, data-driven strategies. By leveraging scalable data technologies, I help cross-functional teams turn insights into actions and drive meaningful business outcomes. Prior to my transition into Data Science, I spent 6 years in Digital Advertising, focused on Ad Technology and Attribution Reporting. As a Senior Account Manager, I partnered with sell-side clients to optimize ad revenue, and with buy-side clients to measure and understand advertising effectiveness. Earlier in my career, I spent 12 years investing in the global capital markets, with a specializing on quantitative strategies. Twice, I was part of a team that managed portfolios with over $500 million in assets under management. My work spanned portfolio management, investment research, client communications, and consulting with entrepreneurs during the early stages of their businesses. I have blogged about my Grandma, traveled extensively, and am a soccer obsessive. True story: I won a Moth GrandSLAM in Seattle, and had a story called "What are you doing over there?" play on the Moth Radio Hour, The Moth podcast, and featured on The Good News Network.
Operations Intelligence (3/25 to present) Architected and deployed scalable data pipelines supporting reporting for Green Apron Service and Starting Five pilot program. Created automated reporting for COO-level meetings and Board of Directors presentations, simplifying complex operational data used for executive-ready insights. Staffing Forecasting (12/20 - 2/25) Provided strategic guidance for existing and new in-store promotional events, offering multi-scenario modeling and difference-of-difference methodologies to provide a range of transaction lift outcomes. Partnered with Finance and Operations teams in US and Canada to deliver fiscal year and quarterly planning forecasts, directly affecting staffing allocation strategy. Part of the team that developed, built, and launched Starbucks' internalized Staffing Model.
fastbal is FAntasy Soccer Team By ALgorithm. I've automated team selections for mlssoccer.com's Fantasy Soccer game using machine learning for predictions and linear programming for optimizing the team. Data was collected from the mlssoccer.com fantasy website using Selenium and BeautifulSoup, the neural network predictive model was built with PyTorch, and the PuLP library was used for the linear programming needed to choose the optimal team construction. All of these libraries are written in python. Results for the 2020 season are shared at the page's github: https://github.com/rsherer/fastbal/blob/master/2020_season_standings.md
Karat, which conducts technical interviews for software-driven companies, sought to reduce post-interview workload for their Interview Engineers. Created an ensemble classification model which generated probabilities of various style labels for code from technical interviews. My model combined tf-idf with a Stochastic Gradient Descent Classifier to provide results delivered through a command-line script.
Selected from a cohort of 17 graduates to support the education team for a Data Science Immersive (DSI) cohort. Taught a Python Fundamentals course, covering numeric types, data structures, object-oriented programming, pandas, numpy, and linear regression. Advised DSI students on their capstone presentations, including guidance on proposals, data acquisition, data cleaning, and modeling.
Certificate program with foundation in python, learning both frequentist and bayesian statistics, machine learning, natural language processing, recommender systems, and neural networks. Case studies used predictive and classifier models to find equipment sales prices, app customer churn prediction, movie recommender system, and fraud detection. For the capstone project, I built a sales prediction model for Evergreens, a regional restaurant chain. The goal was to provide a tool to aid decisions around staffing allocations.
For my capstone project at Galvanize, I built a sales prediction model to guide decisions around staffing allocations. EDA revealed seasonal and day-of-week revenue trends, and tuned a Gradient Boosting model with weather forecasts as the prediction inputs. Utilized time series cross validations on sales data and historic weather data from METAR. Model accurate to within 3% of actual daily sales.