cv

Research Intern

Waymo

Improving robustness of motion forecasting model via closed-loop reinforcement learning fine-tuning.

• Designed large motion forecasting architectures that reduce catastrophic forgetting during RL fine-tuning for safety-critical autonomous planning.
• Built a JAX-based closed-loop simulator with procedural OOD scenario generation, enabling scalable and closed-loop RLFT under distribution shift.

ML Lead, Technical Advisor

Vulcan Augmetics

Developing robust human–machine interfaces through embedded biosignal learning and real-time intent decoding on edge devices.

• Led R&D of embedded bio-signal systems for real-time muscle activity decoding, integrating signal processing and machine learning on edge devices.
• Designed learning pipelines for EMG-based intent recognition with low-latency inference under hardware constraints.
• Advised on end-to-end system design spanning sensors, embedded firmware, and user-facing mobile applications.

Electrical and Electronic Engineer Intern

Bosch

Building backend, monitoring, and IoT data interfaces for an agricultural disease prediction system.

• Developed an OCR-based pipeline for automated SIM card serial number recognition in manufacturing workflows.
• Built a real-time sensor monitoring and reporting system managing over 2,000 sensors with automated fault detection.
• Implemented REST-based data crawler with DFS for data collection and processing.

Embedded Software Engineer Intern

Aubot

Developing embedded control and firmware systems for teleoperated mobile manipulators.

• Developed and deployed control stack for an 8-DOF robotic arm and omnidirectional mobile base with obstacle detection and collision avoidance.
• Implemented sensor-driven motion control and teleoperation for a telepresence robot.

Application Engineer Intern

National Instruments

Prototyping and validating robotic control systems using industrial instrumentation and rapid development platforms.

• Prototyped a self-balancing robot and a 3-DOF robotic arm using NI products.
• Support clients’ projects using NI LabVIEW, Multisim, Ultiboard, and vision toolkits.

Doctor of Philosophy

Princeton University

Electrical and Computer Engineering

Bachelor of Engineering

RMIT University Vietnam

Electrical and Electronics Engineering (Honours First Class)

Outstanding Teaching Assistant Award

Princeton University

First Year Fellowship in Natural Sciences and Engineering

Princeton University

Outstanding Graduate of Degree Program

RMIT University Vietnam

RMIT Certificate of Achievement

RMIT University Vietnam

Best Concept Award

Hackaday Prize

Best Paper Award

BDCloud

From refusal to recovery: A control-theoretic approach to generative AI guardrails

The International Association for Safe & Ethical AI (IASEAI 2026)

Proposes a control-theoretic framework for generative AI guardrails that moves beyond static refusal, enabling recovery and safe continuation through dynamic constraints and feedback-based interventions.

Provably optimal reinforcement learning under safety filtering

The International Association for Safe & Ethical AI (IASEAI 2026)

Develops a theoretical framework establishing optimality guarantees for reinforcement learning under safety filters, showing that appropriately designed filters can enforce safety without sacrificing asymptotic performance.

MAGICS: Adversarial RL with minimax actors guided by implicit critic Stackelberg structure

Workshop on the Algorithmic Foundations of Robotics (WAFR 2024)

Presents an adversarial reinforcement learning algorithm that leverages a Stackelberg game structure with implicit critics, yielding stable and convergent training for safety-oriented robot control policies.

Gameplay filters: Robust zero-shot safety through adversarial imagination

8th Annual Conference on Robot Learning (CoRL 2024)

Introduces Gameplay Filters, a safety framework that leverages adversarial imagination to anticipate and mitigate unsafe behaviors at deployment time, enabling robust zero-shot safety for reinforcement learning policies.

Sim-to-lab-to-real: Safe reinforcement learning with shielding and generalization guarantees

Artificial Intelligence (Journal)

Introduces a sim-to-lab-to-real reinforcement learning framework that integrates safety shielding with formal generalization guarantees, enabling safe transfer of learned policies from simulation to real robotic systems.

Isaacs: Iterative soft adversarial actor-critic for safety

5th Annual Learning for Dynamics and Control Conference (L4DC 2023)

Presents ISAACS, an adversarial actor-critic algorithm that models safety as a soft minimax game between control and disturbance policies, enabling stable learning of safety-aware behaviors in continuous control settings.

Back to the future: Efficient, time-consistent solutions in reach-avoid games

2022 IEEE International Conference on Robotics and Automation (ICRA)

Develops efficient and time-consistent solution methods for reach-avoid differential games, addressing fundamental challenges in safety-critical planning and scalable robotic decision-making.

A hybrid indoor localization system running ensemble machine learning

2018 IEEE ISPA/IUCC/BDCloud/SocialCom/SustainCom

Introduces a hybrid indoor localization system that combines sensor fusion with ensemble machine learning techniques to improve robustness and accuracy in complex indoor environments.

DARPA Learning Introspective Control (LINC) program

Developed introspective safety mechanisms for adaptive robot control under uncertainty, field-tested on a hybrid track-based robot and crane robot

Robust learning-based safety filter on quadruped robot for locomotion

Developed zero-shot safety filter using adversarial imagination, deployed on a quadruped robot to prevent falls and unsafe actions during locomotion.