Demos

Generative Dense Chains (GDC)

The last chapter of my thesis is on Generative Dense Chains. These are created via a one-to-one transform of a time series into a Hidden Markov Model with one state for each element in the time series (plus an additional absorbing state). I developed these models to measure distances between arbitrary time series, however they also preform as excellent forecasting models over both time series and text, allowing for full explainability with each prediction. Scroll down for a demo with your own text.

LBM thermofluid sim

This was a project to test the limits of new llms for building a visual. I provided the equations for the fluid-transport, heat-transport, and coriolis effect, then had o3-mini-high handle the visual. Careful about editing the parameters, it can be prone to instability under different conditions. This simulation became the basis for a dataset generated for the 2025 datahack on weather forecasting for mlds, which you can check out here

Page updated

Google Sites

Report abuse