My current research focuses on biologically inspired computer vision and perception. My work is in understanding the computational circuits that underlie robustness, adaptability and modeling.
I am a Principal Research Scientist at NAVWAR in San Diego, where I founded and direct the NeuroML and NeuroVision labs. I am completing my Ph.D. in Data Science at UC San Diego as a DoD SMART Scholar, advised by Dr. Virginia de Sa in Cognitive Science and Prof. Gary Cottrell in CSE. Previously I received my M.S. in Computer Science from Georgia Tech, where I was advised by Prof. David Joyner and Prof. Ashok Goel, and my B.S. in Electrical Engineering from the University of Rochester.
NeurIPS 2026 · Evaluations and Datasets Track
On the 64,420 non-tie triplets, cosine distance on a frozen self-supervised backbone reaches 74.1% accuracy without training, LoRA fine-tuning reaches 77–79%, and the DreamSim ensemble reaches 80.7%. A split-half reproducibility estimate from the triplet vote counts shows that majority labels reproduce only 70.1% of the time overall.
Conference on Cognitive Computational Neuroscience (CCN) · 2026
Gradient-guided pixel perturbations shift predictions by ~0.56, more than eight times the person-presence effect, indicating that these models rely on pixel-level statistics rather than the scene-level processing that underlies human memorability.
arXiv:2310.07941 · 2023
Our goal was to classify cognitive events from EEG signals while subjects simultaneously walked on a treadmill, sometimes while carrying a rucksack equivalent to 40% of their body weight. To our knowledge, these results are the first documented implementation of a DNN for the classification of cognitive neural state during dual-task walking.
IEEE Signal Processing in Medicine and Biology (SPMB) · 2021 · pp. 1–5
We evaluated an approach to address this limitation by modifying a U-Net convolutional neural network to extract blink artifacts from EEG signals contaminated with electrooculography signal. Our model removed ocular artifacts with a higher accuracy compared to the industry standard ICA algorithm.
Proc. SPIE 11733, Geospatial Informatics XI · 2021 · 1173305
We address a lack of class coherency across datasets at varying spatial resolutions by introducing a large 42-class synthetic dataset, Maritime Vessels at Varying Resolutions (MVVR-42). We explore the effect of spatial resolution on performance in ship detection tasks.