Multiscale Bayesian framework bridging 5-species oral biofilm ecology and 3D FEM stress analysis via TMCMC, DeepONet surrogate (~100x speedup), and GNN-informed priors. 28x stiffness range, basin sensitivity, gradient-based NUTS sampling.

GPU-accelerated Bayesian inference for Rust, browser, Python, and R

🔬 Autonomous Bayesian scientific agent — MCMC analysis, uncertainty quantification, and permanent decentralized archiving. The reproducibility layer for DeSci.

An MVP system for feature engineering and causal discovery for pump equipment. Generates 22 features from 90-day measurement data and discovers causal relationships to anomaly prediction using LiNGAM (Linear Non-Gaussian Acyclic Model).