Optimize Deere’s Feature Store to easily aggregate 13 petabytes of agronomic and embedded system data gathered from onboard agriculture equipment using Spark to serve to 100+ data analystsSimplified Deere’s weather API services using AWS Chalice, API Gateway, Lambda, and Terraform, creating a low-latency API for 100,000's of customers in production systemsDeveloped GIS data processing pipelines (Spark, S3) to support brand sustainability goals, providing end users with personalized information on nitrogen-use efficiency metrics, increasing subsciption benefitsImproved feature store API, allowing internal customers to efficiently utilize petabytes of data for use in production systems used by over 20 internal teams building more than 50 productsLed my team in using Sentinel-2 and PlanetLab satellite imagery to provide NDVI (vegetation index) metrics in customer fields, leading to a possible increase in recurring revenue by $117 millionBuilt algorithms to dynamically adjust MySQL Database metadata in a low-latency feature store for consumer agricultural applications, enhancing data efficiency and qualityLed a project to forecast used inventory levels at the John Deere dealer level using neural networks and traditional statistical methods to influence $37 million incentives given to customersProvided SQL tools to our pricing specialists and other business stakeholders to query data in databases, converting them from Excel spreadsheets to curated datasets in our enterprise Data LakeCollaborated with cross-functional teams, including engineers and non-technical stakeholders, for the successful creation and deployment of deep learning models using Tensorflow and MLFlow

Mentor beginner programming students ranging from middle school to high school through the creation of interactive projects coded in Python; teach and explain basic programming logic and Python language syntax as well as valuable coding practices, including debugging techniques and code organization.

• Updated and improved an algorithm for optimal commission pricing in commercial real estate markets nationally, highlighting a $10 million revenue gap• Automated data analysis for market shares for over 1000 markets. Moved from a process that took an employee days to a process that was automatic and took five minutes.• Created data visualizations for linear regressions and their confidence intervals to identify under-performing and over-performing brokers for every U.S. broker team.• Used machine learning techniques like one-hot encoding and binary variables to simplify model creation pipeline