Geometry of Poker

Geometry of Poker is a research‑grade project where I build an interactive 3D visualization of the entire Texas Hold’em state space, treating poker situations as points in a high‑dimensional mathematical space and then embedding them into three dimensions using PCA and UMAP. Instead of hand‑drawing graphics, I define strategically meaningful numeric features for each state (such as equity, vulnerability, and board texture) using a published poker engine, and those feature vectors are what drive the geometry of the visualization. The codebase is a TypeScript/Python monorepo: a Next.js and React Three Fiber web app renders a GPU‑accelerated point cloud in the browser, while a separate Python pipeline handles dataset generation, normalization, dimensionality reduction, and clustering with libraries like scikit‑learn, umap‑learn, and hdbscan. The system is artifact‑driven, meaning the viewer consumes precomputed binary artifacts produced by the pipeline, which makes it suitable for product