Data Science and Applied Mathematics
Data Scientist Advisor in drilling at Continental Resources, Oklahoma City.
I trained in mathematics, spent fifteen years in upstream drilling operations, and returned to school for an M.S. in Data Science. I now build real-time analytical systems on rig sensor data: streaming pipelines, predictive models for performance and equipment failure, and the engineering computations that sit beside them.
Outside that work I write small, narrowly scoped studies. Each one asks a single question, holds everything else fixed, and states what the evidence does and does not support.
Reduces the coin-triangle inversion problem to finding the nearest integer point on the plane k₁ + k₂ + k₃ = n, recovering the minimum ⌊n(n+1)/6⌋ and every optimal configuration. Extends the argument to d-dimensional piles: optimal placements are balanced, tetrahedral piles have a closed form, and the disjoint-corner pattern fails from dimension four.
The Calibration–Power Trade-off Between Parallel Analysis and Spectral Thresholding
A simulation study of false retention and detection power for parallel analysis and Marchenko–Pastur thresholding, using 686,500 synthetic datasets with known latent structure.
An Empirical Study of Target Transformation in Rate-of-Penetration Prediction
A controlled comparison of raw and log1p targets across three geothermal wells and five model families, evaluated on unseen depth intervals.
A decoder-only transformer written from scratch in PyTorch and trained on a laptop, using short stories and exactly verified integer arithmetic. The goal is understanding each part, from tokenizer to decoding, not a research claim. The first 27M-parameter model writes coherent short stories but solves about one arithmetic problem in twenty: it learned the answer format, not place value.
Complexity should be earned by the problem, not assumed by default.