Hi, I'm Jeff

Jeff Keller

Lead Data Scientist at E Inc.

Experienced Data Scientist with strong statistics, technology, and consulting skills. Adept at conveying complex concepts to audiences of varying technical abilities. Expert in modern “full-stack” statistics, relevant technologies, and bringing value directly to business applications.

Jeff lives with his lovely wife Valerie in Georgia, Vermont, where they have resided since 2017.

Fast Learner
Critical Thinker



Principal Data Scientist

Aug 2022 - Present, Burlington, VT

E INC is on a mission to create the best digital auction and retailing platform in the world by connecting the automotive wholesale and retail experiences—two worlds that have historically been kept apart—together.

  • Responsible for scaling the company’s Data Science discipline in an AWS cloud environment
  • Partnered with Engineering and Infrastructure leadership to establish highly reusable patterns for frictionless continual delivery of Data Science products
  • Developed a wholesale vehicle recommendation engine to support Product, Marketing, and Operations use cases
  • Delivered and enhanced a vehicle pricing engine to support business critical financial products
  • Regular advisor to the architecture and implementation of the company’s rapidly growing Data Lake and Data Warehouse

Cox Automotive

Sep 2016 - Aug 2022, Burlington, VT

Cox Automotive is an Atlanta-based business unit of Cox Enterprises, formed in 2014 to consolidate all of Cox’s global automotive businesses, including Kelley Blue Book, Xtime, Autotrader.com, and Manheim.

Lead Data Scientist

Jan 2018 - Aug 2022

  • Invented an enterprise hierarchical Bayesian Markov modeling capability for marketing attribution, budget optimization, customer intervention, and session valuation
  • Developed a joint glmnet + Random Forest model to optimize Search ad campaigns
  • Pioneered cloud-first Data Science ModelOps patterns and infrastructure
  • Deployed and administrated Data Science workbench, publishing, and deployment platforms and environments
  • Mentored a group of 8 to 12 junior and senior Data Scientists
Senior Data Scientist

Sep 2016 - Dec 2017

  • Designed and productionized an enterprise consumer vehicle recommendation engine
  • Built Random Forest models to predict vehicle sales from anonymous website activity
  • Led the Data Science organization’s transition into the cloud (AWS)
  • Created R and Python packages to increase productivity of other Data Scientists


Jun 2010 - Sep 2016, Burlington, VT

RSG applies unmatched research and analytics to inform clients’ strategy and planning, helping organizations make critical decisions with confidence.

Lead Analyst

Jan 2014 - Sep 2016

  • Designed and implemented novel agent-based economic simulation models for government policy and scenario planning use-cases
  • Productionized and integrated embedded machine learning business applications
  • Designed and implemented advanced agent-based micro-, meso-, and macro-scale economic simulation models
  • Developed custom data ETL and warehousing tools
Senior Analyst

Jun 2010 - Dec 2013

  • Developed statistical models of consumer purchase behavior for optimizing prices and product configurations
  • Designed research experiments to quantify consumer preferences and willingness-to-pay
  • Segmented and personified customer audiences using latent class clustering techniques


Loot of Lima Companion App
Loot of Lima Companion App
Developer Aug 2020 - Present

Companion web app for the treasure hunt-themed deduction tabletop game, Loot of Lima.

Developer Dec 2018 - Present

A Windows R package for simulating keyboard and mouse inputs.

Developer Dec 2018 - Present

An in-development FLOSS companion app for space and flight simulator games such as Star Citizen. Connect a tablet or other device to your PC to create a Multi-Functional Display (MFD).

SGFD Water Rates
SGFD Water Rates
Developer Aug 2020 - Present

A web application/calculator of the 2021 metered water rate system with tiered pay scales for the community of 190 households in the South Georgia Fire District of Vermont.

Maintainer May 2015 - Sep 2016

Functions for estimating models using a Hierarchical Bayesian (HB) framework. The flexibility comes in allowing the user to specify the likelihood function directly instead of assuming predetermined model structures.


Recent Posts


Best Poster ART 2014

Best Paper TRB

ABJ70 Member (2014-2017)