CV

Research Assistant

Georgia Institute of Technology

Laboratory for Intelligent Decision and Autonomous Robots (LIDAR), Advised by Prof. Ye Zhao

TAPNAV: Humanoid Navigation through Tactile Active Perception

• Introduced a humanoid tactile navigation framework for vision-denied environments, where the robot localizes by actively probing nearby structures.
• Developed a whole-body control policy tracking end-effector poses, contact forces, and footsteps, trained with offline IK data, a virtual spring-damper field, and 3D-LIPM references.
• Built a ROS 2 full-stack pipeline integrating tactile sensing, proprioception, control, and global/local planning.

Graduate Research Assistant

University of California San Diego

Advanced Robotics and Controls Laboratory (ARClab), Advised by Prof. Michael Yip

SteadyTray: Learning Object Balancing Tasks in Humanoid Tray Transport via Residual Reinforcement Learning

• Introduced ReST-RL, a student-teacher residual RL framework that keeps unsecured objects on a tray stable while the humanoid walks and actively rebalances them under disturbances.
• Trained a residual module on top of a base walking policy in Isaac Lab, distilling a teacher with privileged object states into a student that sees only the AprilTag pose.
• Achieved zero-shot sim-to-real transfer on a Unitree G1, carrying diverse objects such as a water-filled wine glass under external pushes.

Dataset-Free Compliant Humanoid Robot Loco-Manipulation with Dynamic Online Posture

• Introduced OCLO, a humanoid loco-manipulation system that takes only target hand poses as input and generates its own whole-body posture, trained without human motion data.
• Implemented whole-body compliance: IK-controlled arms track force-displaced spring-damper references, while the lower body learns to adapt to external upper-body force and motion.
• Enabled the Unitree G1 to stay balanced under end-effector pulls and pushes, where ablations without this force training mostly failed.

Undergraduate Research Assistant

Lehigh University

Seed Lab, Advised by Prof. Ning Li

Methods for Setting Device Specifications for Analog In-Memory Computing Inference

• Introduced a systematic approach to setting device specifications for analog in-memory computing, identifying device parameters that keep DNN accuracy within 1% of floating point.
• Applied hardware-aware training with PyTorch and IBM AIHWKit to make LSTM and ResNet-32 models robust to analog noise before mapping them onto simulated devices.
• Evaluated how PCM device non-idealities (conductance range, drift, programming and read noise) affect accuracy over time, through large-scale simulations on HPC clusters.

Undergraduate Research Assistant

Lehigh University

SwarmsLab, Advised by Prof. David Saldaña

• Built lighter-than-air blimp robots for the “Defend the Republic” swarm competition, from 3D-printed frames to onboard IMU, ultrasonic, and camera sensing.
• Implemented PID control and camera-based blob tracking for the blimps, with ground-station communication for manual and autonomous modes.
• Developed ROS 2 ground robots with AprilTag visual servoing, Kalman-filter SLAM, and A*/RRT/PRM planning.

Undergraduate Research Assistant

Lehigh University

DASH Lab, Advised by Prof. Yu Yang

• Developed a Flutter mobile app that collects air-quality data from portable Bluetooth sensors in areas without monitoring stations, supporting future ML-based prediction.
• Added real-time location tracking, live chart and map visualization, and cloud storage on MongoDB Atlas.

Master of Science

University of California San Diego

Computer Science

• GPA: 3.87/4.0

Bachelor of Science

Lehigh University

Computer Science

• GPA: 3.82/4.0, Graduated with Highest Honors