Dr. Joel Lapin

My name is Joel Lapin and my interests are in deep learning and modeling in the natural sciences. My projects include the development of a foundational model for encoding spectra into latent spaces. These spectrum encodings can then be used in downstream applications, such as denovo sequencing, spectral similarity scoring, etc. For training such a model, I am actively developing self-supervised training methods to avail the model of the vast amounts of publicly available unlabeled data. My other research interests include the application of explainable AI to peptide property predictors. Specifically I have been using SHAP analysis to models for intensity, retention time, and charge state prediction, i.e. primarily models whose input are amino acid sequences. My hope is to use explainable AI to expand scientific understanding of the physical/chemical processes underlying the predictions where first principles understanding is currently lacking.