Brian Neely

I'm an engineer and analyst who likes turning messy data into tools people can use. My background runs from nuclear engineering and project controls to data science and business analytics. I build my own software for data analysis, automation and personal projects, and I'm always interested in hard problems that make a real difference.

A permutation-invariant surrogate for Monte Carlo k_eff: neutron moderation as the dominant failure mode
Brian Neely · working draft, not yet peer reviewed · target: Annals of Nuclear Energy

A Deep Sets neural network trained on 160,000 MCNP criticality calculations of 1–20 fissile spheres predicts k_eff for arbitrary arrangements. Controlled datasets show that neutron moderation, not geometry, is what makes the problem hard, and the model can still safely screen out a share of runs without missing supercritical configurations.

Read the current draft (PDF) — updated as I write; comments welcome at [email protected].

• Data, analytics and operations roles at T-Mobile, Amazon, Puget Sound Energy and Keyence2018 – today
• Los Alamos National Laboratory — project controls and forecasting for a $600M facility project2017
• Vigor Industrial — project engineering on AP1000 reactor module construction2015
• Cambia Health Solutions — data scientist, statistical modeling2014 – 2015

• Oregon State University — M.S. and B.S., Nuclear Engineering2009 – 2016
• Willamette University — Master of Business Administration2013 – 2015
• Six Sigma Green Belt (IIE) · Engineer Intern (passed the FE exam)

neelybd.com is my personal domain. It hosts this page and a few self-hosted tools I've built for my own use (personal finance, health tracking and document editing). The tools are private and require a login.

Email: [email protected]
LinkedIn: linkedin.com/in/neelybd