Samir Darouich

Structurally accurate molecules often violate basic physics at stationary points, but penalizing energy curvature via ML-predicted Hessians slashes force deviations by 63% without expensive quantum chemistry.

Spectral fingerprints can differentiate between unique molecular structures with identical 2D connectivity, revolutionizing how we assess chemical similarity.

A unified platform for molecular machine learning that supports 100 elements and incorporates uncertainty quantification could democratize access to advanced chemical property predictions.

ARIA reallocates training focus to areas of persistent misalignment, leading to significant performance improvements in unseen conditions.