Eric Hall Email & Phone Number

Minneapolis, Minnesota, Us

Minneapolis, Minnesota, Us

Product owner leading a team of General Mills employees and external consultants to migrate all of our data science team's code base from our on-premise, Hadoop based infrastructure to Google Cloud Platform (GCP). Responsibilities include the on-time and on-budget delivery of our codebase, which impacts analytic decisions made in many areas of the company including sourcing, supply chain, marketing, and sales. To do this I am in charge of scoping the existing projects, architecting the project to suit GCP, setting strategy and direction for the Data Engineers and Data Scientists on the team, reviewing and validating the migrated code, and finally delivering back to the code owner.

Minneapolis, Minnesota, Us

Provided data science expertise for the company's global ecommerce team. A difficult problem for our company is that we do not own our point of sale data, so to understand our ecommerce business we must gather data from many disparate sources, clean them and bring them together in order to drive business understanding. As part of the team, helped to build a new enterprise tool called "the Data Tuner" which allowed semi-automatic matching of new data to existing data to expedite this process. Also assisted in understanding the impact of how a product's search result on a retail website affects its sales at a retailer/weekly/keyword specific level. (2018 - Present) Assisted in building a model to predict the impact of various promotional activity on the sales of a product. The model, built using primarily Impala and Python (pandas & sk-learn), accounts for nearly 100 variables and learns from millions of rows of historical sales data. The model has been integrated into a user friendly website that is used by our sales teams to plan promotional activities for fiscal year 2021 and beyond. (2018 - Present) Lead initiatives to build reusable code packages in python for other data scientists and analytic practitioners at General Mills to use. One package, called gmiEDA, allows the user to calculate and display common statistics and informative aspects of a dataset quickly, without having to write any new code. Another package, called the forecast engine, allows data scientists to build models directly from our data lake in python with simply a configuration file, avoiding having to spend time worrying about data connections, data cleaning and munging, or parameter tuning. This package allows users to quickly get a new project off the ground and iterate on existing models

Chicago, Illinois, Us

Lead research initiatives on automation of creating anomaly detection rules for Uptake's partners in industries like mining, construction, aviation, wind energy, oil and gas. Used concepts from statistics and machine learning to detect different modes, subsystems and other characteristics of client's data. These automation processes improved the on-boarding process for a client from taking days of manual input from our users to a matter of minutes.Was the primary maintainer and developer of an R package used by fellow data scientists and software engineers to rapidly prototype and deploy anomaly detection rules into the Uptake platform. Used concepts of functional programming to allow for rapid algorithmic research and development and concepts from object oriented programming to facilitate the translation of models into production ready code.

Madison, Wi, Us

Proved sample complexity bounds for autoregressive point processes in non-Gaussian noise settings. Sample complexity bounds dictate how much information needs to be gathered to achieve desired levels of statistical confidence in models widely used in the field of neuroscience. Analysis used novel statistical techniques from the fields of statistical inference, point processes and martingale theory, and combined them in novel ways.

Madison, Wi, Us

Proposed and analyzed streaming optimization method for learning network structure in autoregres- sive point processes. Using ideas from the field of online learning, developed methods to estimate pairwise influences of nodes in a network using only event timing data in a streaming fashion. Pre- vious methods all processed information in batch, which is computationally infeasible for large data sets. Applied methods to learn the network structure in applications such as social, epidemiological and neurological networks.

Durham, North Carolina, Us

Worked in the field of Online Convex Programming with a focus on streaming optimization in dynamic environments. Developed algorithms which seek to estimate both the hidden state of a system and the underlying dynamics of the environment. Proved regret bounds for algorithms which incorporate dynamical models, and applied algorithms to applications such as object tracking, compressed sensing, Poisson video reconstruction and network estimation.

Durham, North Carolina, Us

Taught, managed, and graded undergraduate laboratory sections of 12 people (2009-2010) in the introductory signals course. Appointed head lab T.A. for course in Spring 2010 semester. Held office hours, graded papers and occasionally led lectures for class of 40 students in senior level image processing course (Fall 2010) and graduate level digital signal processing course (Spring 2013).

Arlington, Va, Us

Investigated the role of regularization penalties on image processing of telescope data obtained from the Haleakala observatory. Specifically, tested the effects of including sparsity penalties in a multi- frame blind deconvolution framework in order to denoise images corrupted by atmospheric blur. Additionally, created a method to remove background star streaks from a series of telescope image data using sparsity penalties.

Raleigh, North Carolina, Us

Participated in research program in modeling and industrial applied mathematics at North Carolina State University. Studied the concepts of Epidemiology with advisor Dr. Alun L. Lloyd. Used MATLAB to fit a non-linear 2D partial differential equation using numerical techniques to data from the 2009 H1N1 outbreak. Additionally, studied the efficacy of different vaccine dispersion strategies through simulation. Presented research at an on campus research symposium.

Peachtree Corners, Georgia, Us

Took seminars to learn company’s proprietary software, which allows utility companies to observe and manage their resources remotely instead of having to be in the field. Helped to rewrite the manual which teaches clients about the company’s software. Assisted clients by remotely logging into their systems and troubleshooting software.

Doctor Of Philosophy (Ph.D.), Electrical And Computer Engineering

Master Of Science (Ms), Electrical And Computer Engineering

Bachelor Of Science In Engineering (Bse), Electrical And Computer Engineering; Mathematics