RESEARCH
// EXPLORING IDEAS
Academic research, experiments, and technical deep-dives
Research Projects
MOGO: A Fully-Soft Ball-and-Socket Connector for Modular Soft Robotics
CompletedDecember 2025
An extended abstract on the Connector Compliance Paradox: every surveyed modular soft robot joins soft modules with connectors 100 to 5000 times stiffer than they are. MOGO is a single-material TPU ball-and-socket joint at a compliance ratio near 1, where infill engineering gives 25 N insertion, 55 N pull-out, a 2.2 force ratio and more than 500 cycles without visible wear.
TinyPointNeXt: Lightweight Architecture for Real-Time Human Detection in 3D Point Clouds
CompletedMay 2025
A comprehensive study on efficient point cloud processing architectures, demonstrating that strategic model compression and optimized training procedures can achieve 81% parameter reduction while improving classification accuracy for dynamic object segmentation.
Advancing Echocardiography Segmentation with Machine Learning
CompletedDecember 15, 2023
A deep learning approach to echocardiography segmentation, focusing on cardiac structure identification, 3D reconstruction, and clinical impact.