Tom Shafer

My family

Hi, I’m Tom! 👨🏻‍💻👨🏻‍🔬😊

I work as a Principal Data Scientist for Elder Research, with a special interest in Bayesian modeling, interpretable machine learning & applications to fairness, and data science workflows. Before working in data science I completed a Ph.D. in physics at the University of North Carolina, studying radioactive properties of atomic nuclei with supercomputer simulations.

In the other three-quarters of my time I like to read, write computer programs, and play music. My family and I live in Raleigh, NC and are members of the The Summit Church.


Data Science & Machine Learning

I split my time between study—toying with interesting papers, ideas, and techniques that might windup being useful—and practice, working with other data scientists to solve complex problems for our clients. Consulting is an especially interesting place to practice data science, if only because I get exposed to lots of different problems. I’ve gotten to build anti-fraud systems, study how influence spreads across networks, fit large Bayesian multifactor models, and lots of other fun stuff. More →


I’ve played guitars for quite a long time now, and I’ve been fortunate to work with lots of folks including The Summit Church, Lee Hester, Port City Community Church, and others. These days, I’m actively involved with The Summit Church and Summit Worship here in Raleigh-Durham. More →

Physics Research

My graduate physics research focused on radioactive properties of the very heaviest atomic nuclei—nuclei so heavy that they don’t naturally exist on earth. It so happens that these nuclei’s half-lives are an important puzzle piece in the search for the origin of the heavy elements (heavier than iron).

This research has direct bearing on, for example, LIGO’s 2017 detection of a neutron star merger: a process hypothesized to form these heavy elements. Follow-up observations are trying to determine whether this merger event produced the nuclei we’re searching for.

Practically, our group developed software that significantly reduced the computational cost of these half-life calculations (Phys. Rev. C, preprint), and we used supercomputing resources to calculate properties for thousands of nuclei (Phys. Rev. C, preprint). I further extended the method to compute half-lives of nuclei with odd numbers of protons and/or neutrons, which is a challenging technical problem (Phys. Rev. C, preprint). More →