Sam Helmich Email and Phone Number

Leads Data Science efforts for the Marketing Organization as a player-coach and mentor other Data Scientists throughout the company.

My role is to be a player-coach for a team of 10 Data Scientists.As a Leader-Defined and lead process for engagement between Data Science and Product Management which drove development, implementation, and production of customer facing analytical solutions-Worked with Lead User Groups and Product Management teams to identify and prioritize customer facing analytical solutions-Lead a team of 7 in producing customer facing analytical models by breaking full production solutions into component models, delegating model responsibilities to sub-teams, setting deadlines, and monitoring progress on minimum-viable model development using an agile work management approach-Prepared presentations for many different leadership levels for updated on ongoing work, results of key projects, and proposals for new lines of critical workAs a Practitioner-Implemented Datasheets for engineered features, allowing for easier re-use of variables and reduced time including engineered variables in new models.-Implemented Model Cards for developed models, making the strengths and weaknesses of those models easier to understand by both technical and non-technical audiences, and identifying areas for model improvement-Implemented model tracking with MLFlow (within Databricks), cataloguing model and hyperparameter performance across production models-Developed a statistically based benchmarking methodology to demonstrate and flag improvement opportunities for customers-Regular presenter and organizer of a grassroots knowledge sharing session across all analytics practitioners at John Deere.-Developed framework for criterion-based pruning of convolutional neural networks and implemented it in python-Built predictive models for core economic and performance metrics for farmers from a dataset of more than 3 trillion rows of data using Spark, AWS, and Databricks

My role was to approach the vast data reserves from John Deere Agronomic & Telematics data and demonstrate the value of utilizing them for customer-facing analytics. I worked with a Data Wrangler and a Project Manager as a team of 3 to develop models that could easily speak to customers and leadership. These models lead to an expansion of the data infrastructure for analytics and the size and scope of the data science team.- Developed a deep convolutional neural network in Python using Keras and Tensorflow to predict yield and grain moisture, leading to a broad expansion of Deere’s analytical warehousing and expansion of the product data science team from 3 to 10- Performed geospatial statistical analysis on field trials to estimate differences in crop performance and economic impact- Worked with Agronomists, Hydrologists, and other SME’s to develop useful data science features and models to reduce variability across agronomic analysis and to facilitate more effective field trials- Developed and presented on outlines for critical analytical infrastructure- Won company-wide hackathon for predicting default rates on loans

My role was to support forecasting and demand planning for global Agriculture and Turf complete goods. This involved gathering and aggregating data, analyzing data, and supporting ad-hoc requests for data & modeling from other groups within Order Fulfillment.- Supported the adoption of an R Shiny server through alpha testing and producing multiple prototype applications and templates, leading to widespread adoption throughout the company for web-based model interfaces.- Wrote R program to forecast 1500 products in 4 global regions and a Shiny frontend to distribute interactive analysis to hundreds of stakeholders on a weekly basis- Used lifespan analysis to predict customer churn and machine failures- Developed techniques to predict customer activity around the release of new models and features- Demonstrated value of shiny servers to IT leadership and created prototypes and training materials to onboard other users to the server- Developed analysis to demonstrate value of end-to-end tracking of supply chain costs through effective linkage of critical databases- Developed training content and presented to other analytics practitioners to bring cutting edge technologies and techniques into the mainstream at John Deere- Developed automated SQL queries to reduce data assembly time from 8 hours per week to a few seconds.

Determining usefulness of forecasting variables, production of software related to simplifying business decisions, writing a core forecasting engine for forecasting complete good sales, statistical consulting, and simulating customer buying habits, among other statistical and technical roles.

Created program that analyzed performance goals based on a variety of variables.