About

About Me...

I am currently the Lead Data Scientist at Redbird Analytics and heavily involved in building out maximally deterministic LLM solutions for data transformations. I’m working on understanding the best possible marriage of LLM’s flexibility and nuance with deterministic workflows that can ensure that hallucinations are minimized while still driving flexible and powerful solutions.

I’ve previously led teams of data scientists at Lowe’s Home Improvement, where I was particularly involved in demand forecasting, route optimization, and staffing optimization. I’m especially interested in large scale optimization problems and have done significant work in understanding the best ways to deploy large-scale analytical workloads like multi-hundred-million SKU forecasts and million+ route optimizations. I find joy in figuring out how to properly scale these types of machine learning and AI-based tools to maximize performance while not running a giant cloud compute bill.

In a previous life, I was a postdoctoral research associate at the University of Illinois at Chicago studying the behavior of an incredibly hot and dense state of matter known as the Quark-Gluon Plasma (QGP).  We have observed evidence of QGP production at both the Relativistic Heavy Ion Collider at Brookhaven National Lab, as well as at the CERN Large Hadron Collider. I was a member of the CMS Collaboration at the CERN LHC and the STAR collaboration at Brookhaven National Laboratory.

As part of my research (and outside of it) I'm very interested in computational methodologies, particularly applications of machine learning and "big data" analytics.  I've been using some simple machine learning techniques including Boosted Decision Trees and Neural Networks in order to optimize some of the criteria used to identify certain subprocesses in our collision data from the LHC.  Outside of physics, I'm working on some demo projects involving maximizing efficiency of a 3x3 road grid, where each intersection has a stoplight.  Though stoplights are not the best solution to a densely populated grid, maximization of such structures can dramatically increase vehicle throughput with almost no cost and no construction.

Feel free to checkout my github repository for the raw code to some of my projects, or have a look at some of the demos on the Projects page.