Efficient Autonomous Navigation of a Quadruped Robot in Underground Mines on Edge Hardware
Pre-print · 2026
Runs entirely on a 40 W Intel NUC without a GPU or network connectivity, achieving 100% success across 20 underground field trials.
gradient descending through life 📈 📉
I am the Chief Scientist in Robotics & AI at TouchTronix Robotics, where I lead efforts in vision+tactile data collection and robot policy learning.
Previously, I was a Post-Doctoral Fellow at Missouri S&T, where I used the Spot quadruped platform to develop autonomous systems for miner search and rescue missions. Under the guidance of Dr. Kwame Awuah-Offei, this research integrated autonomous navigation, digital twins, and computer vision. During this period, I also served as a Part-Time Robotic Consultant with Ameren's Innovation Implementation and Strategy team, led by Alex Rojas. There, I worked on deploying Spot for autonomous substation inspections, including step-voltage measurements, to assist field workers. The shared quadruped platform closely connected my research and industry work.
I earned my Ph.D. from University of Missouri - Columbia under the supervision of Dr. Gui DeSouza at the Vision Guided Intelligent Robotics Laboratory. My doctoral research, supported by the NIH, pioneered machine learning applications for voice pathology, culminating in publications that bridge the fields of engineering and clinical science.

Pre-print · 2026
Runs entirely on a 40 W Intel NUC without a GPU or network connectivity, achieving 100% success across 20 underground field trials.
IEEE Congress on Evolutionary Computation (CEC) · 2023
Uses a genetic algorithm to adapt feature vectors, improving SVM generalization while reducing correlation with confounding factors in sEMG vocal-fatigue classification.
Applied Sciences · 2021
Examines data imbalance, signal normalization, and VFI-based labeling for machine-learning detection of vocal fatigue using sEMG data from 88 participants.