Led a cross-functional team to detect Ganoderma disease across a 20-million palm-tree plantation. Demonstrated that high-sunlight satellite images can be cost-effective with near-infrared signals. Built a three-stage AI pipeline (tree region detection, geospatial feature extraction, and k-means clustering) leveraging multispectral bands, atmospheric correction, cloud/shadow masking, and vegetation indices (NDVI, EVI, NDRE). Achieved a 40% improvement in detection accuracy and eliminated 80% of manual field surveying. Implemented CI/CD infrastructure and a real-time operational dashboard.

Machine Learning Researcher at Ehm Global

January 1, 2020 - December 31, 2020

Optimized the perception stack for a semi-autonomous E-Ferry; developed computer-vision models for path detection, obstacle classification, and navigation decision-making using SVMs and classical CV methods. Led end-to-end data pipeline (collection, preprocessing, analytics) and implemented a real-time decision matrix for autonomous route planning and safety handling, reducing operator intervention by 55% and increasing path-planning reliability by 40%.

Robotics Engineer at iRadar

January 1, 2016 - December 31, 2018

Built a real-time motion-tracking system for on-court tennis analytics using a low-cost vision setup with OpenCV. Designed a full perception pipeline (athlete detection, motion segmentation, trajectory estimation) tuned for outdoor lighting, occlusion, and rapid directional changes. In 2018, partnered with R&D and embedded-systems teams to integrate the ML stack with multi-phase motor control, enhancing autonomous-mode stability for EV taxi, bus, and ferry platforms. Delivered contributions across the full SDLC, including requirements, design, implementation, testing, and maintenance.