PerTurbo
PerTurbo is a scalable Bayesian analysis framework for large-scale Perturb-seq screens, with a GPU-accelerated implementation of the probabilistic model using PyTorch, Pyro, and Lightning.
See the associated manuscript, A scalable Bayesian analysis method for large single cell CRISPR screens (in preparation).
Logan is a Ph.D. candidate in Biomedical Informatics at Harvard Medical School building probabilistic machine learning systems that reason over large-scale biological data. His work in the Pinello Lab develops Bayesian & deep learning models — spanning representation learning, generative modeling, and causal inference — to turn high-throughput single-cell experiments into testable hypotheses about biological mechanism, with the broader goal of building computational systems that can reason under uncertainty and close the loop with experimental biology.
Outside of lab, he enjoys training for marathons (and the occasional IRONMAN triathlon) and bikepacking around Vermont.