I am interested in working on making robots function reliably in the real world. My work focuses on system integration, sensor fusion, and optimizing software for edge computing constraints.
ROS2 Stack for a Mobile Fleet — Lab Infrastructure
Contributed to the development of a software stack for a fleet of 8 Clearpath mobile platforms.
Key Contributions: *
State Estimation: Implemented the sensor fusion pipeline (EKF), integrating Visual-Inertial Odometry (VIO) and Ultra-Wideband (UWB) to achieve drift-free localization in GPS-denied environments. *
System Integration: Configured NVIDIA Jetson compute units for on-board autonomy and integrated Kinova Gen3 Lite arms for mobile manipulation. *
Deployment: Helped establish a containerized ROS2 workflow to ensure consistent performance across all agents.
DESSAIM: Decentralized Swarm Coordination — Best Video Award, ICRA 2025 - Field Robotics Workshop paper ICRA 2024
Collaborated on a multi-agent system comprising micro-robots, indoor drones, and mobile manipulators for live performances. I worked on the engineering team responsible for the system's decentralized control .
Awarded Best Video at ICRA 2025. |
Project Page
Benchmarking Open-Vocabulary Models for Construction — ISARC 2025
Contributed to a study evaluating Vision-Language Models (VLMs) for detecting MEP (Mechanical, Electrical, Plumbing) components. I focused on the practical deployment and benchmarking of these models on a mobile "Journeybot" platform.
Key Contributions: *
Data Collection: Managed the on-site data acquisition and dataset curation in high-clutter construction environments. *
Benchmarking: Compared the inference speed and accuracy of fine-tuned YOLO11 Nano against zero-shot models (Grounding-SAM2, Grounding DINO). *
Results: The study highlighted that while open-vocabulary models generalize well, supervised lightweight models (YOLO) remain superior for real-time edge deployment. Paper:
arXiv
Spectral Analysis for Bio-Inspired Sensing — Agricultural Robotics
Collaborated on a study investigating near-ultraviolet (UV) sensing for agricultural monitoring. I assisted in the data collection and analysis to determine if UV reflectance could serve as a viable marker for crop health.
Key Contributions: *
Field Measurement: Performed in-situ spectral reflectance measurements of strawberry cultivars. *
Analysis: Applied pollinator-vision models to translate raw spectral data into contrast metrics, identifying features overlooked in standard RGB/IR imagery. Paper:
DOI 
