Machine-learning research and engineering in the contexts of speech recognition, learning with noisy labels, and autonomous driving.Co-author of 3 patent applications.

Machine-learning research and engineering in the context of video compression

First author of 5 peer-reviewed publications and 1 invited book chapter.Co-instructed 5 courses and TA'd 2 courses, 6 of which focused on machine learning / probabilistic graphical models / deep learning.Selected Project: Developed and instructed first instance of CS 382: Deep Learning Lab (spring, 2018).Selected Project: Co-developed state-of-the-art models and methods for localizing and classifying surgical activities from hand-motion data.Selected Project: Co-developed models and methods for leveraging unlabeled surgical-motion data, leading to the discovery of surgical activities and state-of-the-art performance for motion-based search.

My team was particularly interested in Adaptive Computation Time -- a technique for learning to dynamically adjust computation in recurrent neural networks based on the difficulty of the current task. In my first project, I implemented this paper from scratch in PyTorch, matched the paper's results, and disseminated learnings with the team.In my second project, I developed and patented a new, simplified method for learning and navigating a map from unlabeled video alone (without any classical components such as feature detection, feature description, etc.).

Developed and published algorithms for detecting chemical warfare agents in long-wave hyperspectral imagery. Additionally, led a large project to evaluate radar performance in real-world conditions (from data collection in the vast desert to implementing data analysis and documenting conclusions).

During my MS, I developed new false-alarm-mitigation algorithms for detecting sub-pixel objects in short-wave hyperspectral imagery.During my BS, I developed a new method for obtaining the size distribution over a collection of nanoparticles, using only thermomagnetic measurements from a superconducting quantum interference device (SQUID).