This project applied Reproducing Kernel Hilbert Space (RKHS) methods to financial modeling, using kernel-based approaches to capture nonlinear structure in market data. The team designed a multi-layer architecture combining fast kernels (limit order book, VPIN, Kyle's Lambda, Hawkes Process, VannCharm) and slow kernels (Variance Risk Premium, Macro Motion, Sentiment via FinBERT), with tensor products used to model fast/slow interactions. Individual kernel out-of-sample correlations ranged from −0.016 to +0.023, confirming architectural independence across signal layers. Backtesting surfaced real signals in limit order book flow, short interest, and VPIN. Next steps include cross-asset expansion, a broader data universe, and live deployment refinement.