Autonomous robot designed for inspection and insulation of tapered power distribution poles. Features modular, adaptive gripping and climbing mechanisms capable of handling varying pole diameters. Integrates vision modules for autonomous pole detection and navigation, with complete mechanical design in SolidWorks and rapid-prototype fabrication via 3D printing.

Real-time criminal activity and weapons detection using vLLM, served via Flask and deployed on Jetson Nano. Deployed across two live pilot projects: Mall Security Surveillance and Shelter Home Monitoring.

An autonomous robot that navigates through agricultural fields using visual servoing to detect crops and weeds, then precisely targets agrochemical spray to minimise toxic chemical usage. Eliminates broadcast spraying and reduces farmer exposure to hazardous fumes through full autonomy.

A solar-powered IoT smart trap that autonomously monitors fruit fly infestations using computer vision. Combines a Raspberry Pi 4 + Intel NCS2 VPU to run real-time YOLO inference on-device — replacing manual scouting with always-on, low-cost automation. Best model (YOLOv5n) achieved 92.4% mAP while consuming just 0.275 J per frame.

A camera-based perception system that detects and classifies different object types in real time. Once an object is identified, the system commands the first robotic arm to pick it up and hand it off to a second arm — passing along the object's class so the second robot can sort it into the correct destination automatically. Demonstrates coordinated multi-robot handoff driven entirely by live vision.

A purpose-built end-effector for the Dobot CR5 robotic arm that automates assembly of modular, tapered pipe sections. The worm-gear-driven jaws reach inside each pipe to grip it from the interior, then rotate to thread it directly into the previous section — repeating the pick, grip, and screw sequence to build a continuous run without manual intervention. Fully modeled and validated in SolidWorks.