The SlothBot
A solar-powered, wire-traversing robot for environmental monitoring, deployed at the Atlanta Botanical Garden.
I work on learning-based planning and autonomous decision-making. My interests include imitation and reinforcement learning, uncertainty-aware autonomy, and control systems.
I earned my PhD in Robotics at Georgia Tech, advised by Magnus Egerstedt and Zsolt Kira. My research focused on safe learning and data-driven control for long-duration autonomy.
During my time at Georgia Tech, I also worked on the SlothBot and the Robotarium.
Experience
Zoox · Autonomy Behaviors
Current role: Senior Machine Learning Engineer
Exyn Technologies
Autonomy & Mapping · Internship
Siemens
Planning & Scheduling for Autonomous Teams · Internship
CERN
Superconductors & Devices · Internship
Education
PhD · Robotics
Georgia Institute of Technology
MS · Computer Science
Machine Learning · Georgia Institute of Technology
BS · Mechanical Engineering
Georgia Institute of Technology
Selected projects
The Robotarium
A remotely accessible swarm robotics testbed. My work included learning-compatible safety frameworks and wireless charging for autonomous experiments.
CARGOS
A Compact Active Response Gravity Offload System developed for a NASA capstone project. I worked on the interface and PID feedback loop. Our team received Georgia Tech’s Best Mechanical Engineering Project award.
Safe reinforcement learning
Combining reinforcement learning with robust control barrier functions to guide exploration under uncertain dynamics.
Selected publications
All publications on Scholar2025IEEE ROBOTICS AND AUTOMATION LETTERS
Safe Reinforcement Learning Using Robust Control Barrier Functions
Summary
A differentiable safety layer combines reinforcement learning with robust control barrier functions, guiding exploration when the system dynamics are uncertain.
2022IEEE TRANSACTIONS ON ROBOTICS
Data-Driven Robust Barrier Functions for Safe, Long-Term Operation
Summary
Gaussian-process estimates of disturbances inform robust control barrier functions, allowing robotic systems to adapt their safety constraints as they collect data.
2021IEEE INTERNATIONAL CONFERENCE ON ROBOTICS AND AUTOMATION
Data-Driven Adaptive Task Allocation for Heterogeneous Multi-Robot Teams Using Robust Control Barrier Functions
Summary
Robot teams adapt task assignments using data about their capabilities and disturbances, with robust control barrier functions linking allocation and execution.
2020IEEE INTERNATIONAL CONFERENCE ON ROBOTICS AND AUTOMATION
Adaptive Task Allocation for Heterogeneous Multi-Robot Teams with Evolving and Unknown Robot Capabilities
Summary
Adaptive assignment lets a heterogeneous team respond when robot capabilities change or are not fully known in advance.
Selected patents
Probabilistic collision detection systemUS12658051B2 · Zoox · Granted
Machine-learned object relevance for vehicle planningUS12415549B1 · Zoox · Granted
Wire-traversing robot and method of operationUS11866079B2 · Georgia Tech · Granted