JT Laune, PhD

I constructed a historical dataset of Wikipedia edits and used machine learning to predict whether spikes in activity corresponded to a contemporaneous event (as opposed to normal editing behavior). Then, I set up a server to listen to the Wikipedia SSE edit stream and deployed the best-performing model to predict which pages are being edited in response to current events.

• Discovered new behavior in nonlinear systems using analytic/semi-analytic methods on time series data.
• Designed Monte Carlo simulations to assess qualitative outcomes in a nonlinear system over ${\sim}10^6$ periods.
• Developed a PDE domain to solve for a small signal with ${\sim}5{\times}$ higher resolution than previous work.
• Implemented custom boundary conditions in a PDE solver to capture a signal at the ${\sim}10^{-3}$ scale.
• Led three projects that resulted in first-author publications.
• Synthesized outputs from diverse simulation frameworks using custom tools to reach scientific conclusions.

• Designed 3D PDE simulations (${\sim}10^{8}$ cells, $5+$ refinement levels) to resolve features at the ${\sim}10^{-3}$ scale.
• Executed simulations on GPU-accelerated high-performance computing (HPC) facilities.
• Tested new code capabilities for effectiveness and robustness to capture a signal at the ${\sim}10^{-5}$ scale.
• Developed performance portable software with Kokkos in an open-source repository with CI/CD tools.
• Simulated 100+ interacting fluids to reveal new qualitative system behavior.
• Led a project that resulted in a first-author publication.

• Developed a Python library for an image reconstruction algorithm.
• Developed a Python library for data conversion across simulation frameworks.
• Led a project that resulted in an expository paper on applications of probability theory to large graphs.
• Prepared and delivered lectures for cohort-taught courses.