Thirulok Sundar Mohan Rasu

MS Student in Electrical & Computer Engineering

University of Michigan, Ann Arbor

Thirulok Sundar Mohan Rasu

About Me

I am a Master's student in Electrical and Computer Engineering at the University of Michigan, Ann Arbor, where I work as a Research Assistant with Prof. Jeong Joon Park. My research lies at the fascinating intersection of robotics, autonomous vehicles, and computer vision—areas that I believe will fundamentally reshape how machines perceive and interact with the world.

I'm driven by the challenge of enabling robots to truly understand their environments. From developing 4D perception systems to creating robust deepfake detection models, my work spans the spectrum of modern AI and robotics. I'm particularly passionate about leveraging foundational models, 3D/4D reconstruction, and sensor fusion to build autonomous systems that can navigate complex, dynamic environments safely and intelligently.

Before joining UMich, I completed my B.Tech in Electrical Engineering from IIT (ISM) Dhanbad, where I developed a strong foundation in machine learning and robotics. I've had the privilege of working with exceptional researchers at Carnegie Mellon University and the University of Central Florida, and contributed to real-world robotics solutions at Mowito.

Beyond research, I'm an avid reader who enjoys diving into fiction and contemplating big questions about technology's role in society. I'm always excited to discuss ideas, collaborate on challenging problems, or explore new research directions—feel free to reach out!

Research Interests

🤖

Robot Perception

4D reconstruction, scene understanding, and developing perception systems that enable robots to make sense of complex environments

🚗

Autonomous Navigation

Motion planning, localization, SLAM, and sensor fusion for safe and efficient autonomous vehicle navigation

👁️

Computer Vision

3D/4D reconstruction, multi-view geometry, and visual learning for robotic applications

🧠

Foundational Models

Leveraging LLMs and VLMs for task planning, human-robot interaction, and multimodal understanding

Recent Updates

Dec 2024

Paper "Towards Content-Agnostic Deepfake Speech Detection with Multi-TTS" accepted to ICASSP 2025! 🎉

Aug 2024

Started MS in ECE at University of Michigan with a perfect 4.0 GPA

Jan 2024

Joined Carnegie Mellon University as a Research Intern working on deepfake audio detection

Mar 2023

Started Robotics Software Internship at Mowito, working on vision solutions for warehouse robots

Experience

Research Assistant

University of Michigan, Ann Arbor

Aug 2025 – Present

Working with Prof. Jeong Joon Park on creating 4D datasets for robot perception and learning, designing data collection pipelines and evaluation protocols.

Research Intern

Carnegie Mellon University

Jan 2024 – Sep 2025

Worked with Dr. Arun Balajee Vasudevan on content-agnostic deepfake speech detection. Developed benchmark datasets using multiple TTS models and trained audio-language detection models. Published paper to ICASSP 2025.

Audio ML Deepfake Detection PyTorch

Research Intern

University of Central Florida - CRCV Lab

Feb 2024 – Dec 2024

Worked with Prof. Yogesh Rawat on mitigating activity hallucinations in Video-LLMs. Created annotated datasets and fine-grained evaluation metrics for robust activity understanding.

Video-LLMs Computer Vision Dataset Creation

Robotics Software Intern

Mowito

Mar 2023 – Jul 2023

Implemented perceptual-hashing for dataset deduplication and built production-ready dataset classes. Developed order-completion time estimation and clustering algorithms for warehouse optimization.

Robotics Computer Vision Production Systems

Featured Projects

Robot Localization with Kalman & Particle Filters

Implemented and compared Kalman and Particle filters on PR2 robot under five different scenarios, studying their advantages and limitations

View on GitHub →

Autonomous Mobile Robot Navigation

Complete autonomous navigation system using ROS2 and Nav2 stack with real-time lidar-based obstacle detection and collision avoidance

View on GitHub →

Challenges in Multi-view 3D Scene Reconstruction

Analyzed failure cases of MASt3R for multi-view reconstruction under challenging conditions and documented limitations

Read Paper →