Summary
My name is Craig Fouts, and I am an enthusiastic
scientist/engineer interested in building mathematical
descriptions of living systems in the context of
biomedical data. As a computational research engineer in
the SciLifeLab
at Uppsala University, I create statistical models that
characterize morphology in spatial transcriptomics data.
Previously, I worked in the New York Genome Center's
Technology Innovation Lab,
where I helped develop machine learning tools used to
study genomic pathologies in histological spinal cord
sections. Before that, I received my master's in applied
mathematics from Columbia University after studying
computer science and theoretical mathematics at The Ohio
State University. Outside of research, I enjoy inline
skating, playing the violin, and building cool stuff.
Projects
A compilation of my experiments with self-organizing
systems and emergent behavior. Current methods include a
Brownian motion algorithm, a Vicsek active matter model,
and a branching annihilating random walk. Also a test
bed for RunTime, a framework for building real-time
interactive environments within Jupyter notebooks.
The source code for neural particle automata, a
generalization of neural cellular automata to point
clouds using graph convolutional networks. The model is
trained to parameterize a nonlinear recurrence relation
that, given some initial "seed" configuration, converges
to a target configuration using local message
propagation.
Implementations of semantic segmentation algorithms and
spatial topic models applied to point cloud data.
Current methods include variations of latent Dirichlet
allocation, a neural clustering process, and a discrete
variational autoencoder. Also a test bed for PriorLDA,
a nonparametric topic model with encoded spatial priors.
A knowledge-based web application for robotic access and
retrieval using IBM Watson. The application provides an
interface for textual communication with a Webots
simulation of two robots collaboratively operating a
virtual warehouse. The user can submit queries about the
warehouse's contents as well as request items for
retrieval.
A webcam-based utility designed to provide an accessible
mode of computer interaction for people with impaired
mobility. The user can use ASL-inspired hand gestures
for 3D control in Autodesk Fusion as well as general
mouse control in Microsoft Windows. Developed for the
2021 HackOHI/O competition, where it won first place.
Research
Ettore Randazzo, Alexander Mordvintsev, & Craig Fouts.
July 24 - 28, 2023. "Growing Steerable Neural Cellular
Automata." Presented at ALIFE 2023.
Alexander Mordvintsev, Ettore Randazzo, Craig Fouts,
& Eyvind Niklasson. July 10, 2023. "Isotropic Neural
Cellular Automata." Blogpost.
Alexander Mordvintsev, Ettore Randazzo, & Craig Fouts.
July 18 - 22, 2022. "Growing Isotropic Neural Cellular
Automata." Presented at ALIFE 2022.
Connect