Here's what I found! The repo is: ## 📦 **suenot/270-gaussian-process-trading** 🔗 https://github.com/suenot/270-gaussian-process-trading ### What it is A **Gaussian Process trading system implemented in Rust**, applied to **BTCUSDT data from the Bybit exchange**. It's essentially a hands-on implementation of the concepts I described earlier — uncertainty-aware trading with GPs. ### Key features in the repo - **Implements GP regression from scratch** in Rust (not just wrapping a library) - **Three kernel functions**: RBF (Squared Exponential), Matern 5/2, and Periodic — composable via addition/multiplication - **Cholesky decomposition** for numerically stable matrix solves - **Hyperparameter optimization** via log-marginal-likelihood maximization (grid search over log-space) - **Trading applications**: - Price prediction with confidence intervals (long/short signals at 2-sigma thresholds) - **Regime change detection** — a spike in posterior variance signals the market has shifted - Position sizing inversely proportional to uncertainty - Volatility surface modeling ### Notable details - **Language**: Rust (🦀) - **Status**: Public repo, 2 commits, MIT license - **Structure**: The main code lives in a `/rust` folder - **README is extensive** — includes full mathematical derivations (GP posterior, kernel math, LML optimization) - Part of what appears to be a larger book/course (labeled "Chapter 322") ### In context of our earlier discussion This repo directly addresses the practical implementation side of GPs for crypto trading. It tackles the key challenges I mentioned — computational complexity is handled via Cholesky (O(n³) but stable), and the uncertainty quantification is used explicitly for risk management and position sizing. If you wanted to dive into the code, this is a solid, self-contained implementation to study or fork.