I build machine-learning systems for hard physical and biological data: surrogate models for computational fluid dynamics in the SmartDATA Lab, event detection and source separation for neural imaging, and the research software that keeps experiments honest. I am a Ph.D. student in Electrical and Computer Engineering at the University of Florida.
The highest-signal project pages that explain what I am building and why it matters.
Ph.D. research in the SmartDATA Lab: learned surrogates that stand in for expensive CFD solvers, judged on stability, physical consistency, and generalization.
Six phases of preregistered experiments asking whether a kernel-routed persistent memory helps sequence models: no general advantage yet, but the hypothesis has narrowed from dense radial attention to sparse, top-K-routed separable memory.
Pipeline that maps noisy multichannel time-series observations to compact latent trajectories with a Hierarchical Linear Dynamical System (HLDS), then evaluates those trajectories for classification, separability, and stability.
Research platform for zebrafish neural-imaging video: statistical event detection for voltage recordings and a calcium-imaging experiment program that separates activity from background, artifacts, and noise.
Software I build outside the lab: apps with their own public homes and tools for my own work.
A local-first training and progress app for lifters: fast set logging, a coach that drafts one change with its confidence, and a biomechanics research spine beside the app.
A reasoning-mastery platform for rigorous STEM courses: attempt a problem, get a deterministic diagnosis of the exact idea you missed, repair the gap, and prove it in a new context.
Local-first photo and video culling that ranks a shoot, learns my taste, and feeds a capture recipe back to the camera.
The last direction of the loop asked whether a latent trained with the label in the loop reads better than a static projection; its synthetic gate failed because the planted effect was invisible to the input's own trial mean, the recalibrated rerun passed but favoured the label-free twin, and the real-data input gate then stopped the direction.
Citations, code, and the official University of Florida references for this work.