RCaGP is a general-purpose Gaussian Process framework that combines robustness to outliers and approximation-aware uncertainty—two coupled limitations of sparse GPs that are critical to tackle for challenging tasks like high-throughput BayesOpt.