1. Information System 2. Knowledge Management 3. Software Project Management 4. Software Quality and Certification 5. Certified Tester (Foundation level) from Malaysian Software Testing Board (MSTB) in 2010. 6. Certified Global E-commerce Talent (2018) from Ali Baba Business School and Ministry of Education, Malaysia

Associate Professor Ts. Dr. Yusmadi Yah Jusoh Department of Software Engineering and Information System Faculty of Computer Science and Information Technology Universiti Putra Malaysia (UPM) Serdang, 43400 Selangor, MALAYSIA Reference: Edge AI-Driven Multimodal Data Fusion for Autonomous Tea Plantation Management in Resource Constrained Environments Yusmadi Yah Jusoh and Xie Qizhao With the rapid development of artificial intelligence, the fine precision of precision agriculture is changing. However, the small-scale tea plantations in resource-constrained areas have difficulties such as delayed disease discovery, inaccurate irrigation decisions and high technical threshold. This study designs and implements an edge AI-based multimodal data fusion system for autonomous tea plantation management. The system uses Raspberry Pi 4B combined with Google Coral Edge TPU, with a single unit price of 2897 RMB and a power consumption of 13.4w, equipped with RGB cameras, multi-depth soil moisture sensors, temperature and humidity sensors, and illuminance sensors, etc. Microservices architecture built out of five different module which are all interchangeable thanks to functioning made using MQTT. There are three algorithmic breakthroughs. Will first use Quantization aware Training, Structural Pruning, Knowledge Distillation so that my Yolo 8 can be 12.7 mb --> 2.6mb (79.5%) compressed and still have 87.8% accuracy and be 16.8 ms inference – that makes it 5.1x faster. HDR synthesis, Polarizing filters, and color calibration together for an average accuracy of 87.1% across all lighting conditions. Secondly, a 85.5% accuracy of disease risk evaluation is increased to 91.5% with an accuracy of F1 score 91.5% via an attention mechanism based multimodal fusion framework. In temporal alignment there are 86.4% of the samples under 10 second difference and spatial registration at 3.2 cm. Third and last a three level of anomaly detection should be able to reach 100% data safety. Main contributions include: propose edge AI architecture for resource limited; Systematically explore model compression strategies; Design attention-based multimodal fusion framework; build a complete data quality assurance system. Provide feasible technical paths for precision agriculture edge intelligence.  Keywords: Edge Artificial Intelligence; Multimodal Data Fusion; Tea Plantation Management; YOLO Model Compression; Precision Agriculture