Invited Speakers

Prof. Wei-Chang Yeh

National Tsing Hua University, Taiwan

Prof. Wei-Chang Yeh is a Chair Professor (ASPEED, NTHU, and CYCU) in the Department of Industrial Engineering and Engineering Management at National Tsing Hua University (NTHU), Taiwan. He leads the "Tsing Hua² Intelligent Robot Team," a cross-university research collaboration of approximately 50 members spanning NTHU, NTU, and NCKU, which has received 18 international competition awards in autonomous mobile robot navigation, digital twin simulation, and semiconductor vision inspection. Prof. Yeh has authored almost 400 peer-reviewed SCI papers and holds more than 70 patents. He has been named among the World's Top 2% Scientists by Stanford University and Elsevier for five consecutive years (since 2020). His research spans network reliability theory, metaheuristic optimization (including the Simplified Swarm Optimization and BAT algorithmic frameworks he developed), autonomous mobile robot navigation reliability (δ-corridor, DP-BAT, TM-A*), digital twin simulation with NVIDIA Isaac Sim / Omniverse, and AI-driven semiconductor wafer identification.He serves as Associate Editor of IEEE Transactions on Reliability and has organized multiple special issues on AI, reliability, and robotics. Prof. Yeh is a long-term collaborator under the NVIDIA Academic Grant Program, working with NVIDIA research engineers on autonomous mobile robot reliability, digital twin deployment, and local large-language-model infrastructure for academic workflows. He has delivered plenary and invited lectures at international conferences including ICSRS 2025 (Turin) and GTC 2026 (San Jose), and is scheduled for an extended invited lecture series at Korean universities in June 2026. Beyond publication and pedagogy, Prof. Yeh has led multi-year industry-academia projects with ITRI (Industrial Technology Research Institute, Taiwan) on Real2Sim contact-rich manipulation platforms, hospital logistics robotics, and companion robot ecosystems. His current research focuses on the deployment of closed-loop private AI pipelines on edge GPU infrastructure—specifically NVIDIA DGX Spark—for academic knowledge work, encompassing thesis evaluation, automated assessment generation, student weakness analysis, instructor guidance, and multilingual voice-cloned content delivery.

Title:  From Grading to Guidance to Peer Review: Deploying a Closed-Loop Academic AI Pipeline on NVIDIA DGX Spark
Abstract: As large language models transform knowledge-intensive workflows, academic institutions face compounding demands: sensitive data—student theses, assessment responses, voice samples, unpublished peer-review manuscripts—require private, sovereign processing, while shrinking budgets preclude high-end GPU workstations. This talk presents a local-first, closed-loop big data pipeline deployed on NVIDIA DGX Spark (128GB unified memory, Grace-Blackwell architecture), selected over a $10K RTX Pro 6000 workstation to enable cost-effective replication across a 50-member research team spanning three universities. The pipeline integrates five interconnected modules: (1) End-to-End Assessment Pipeline, covering the full quiz lifecycle—automated question generation with per-student variants for integrity, hybrid deterministic/LLM-based scoring, statistical analysis (item difficulty, discrimination index, Cronbach's α reliability), and real-time class-level performance dashboards; (2) Structured Thesis Evaluation, combining deterministic PDF parsers with LLM-based six-dimension analysis, cross-referenced against OpenAlex for novelty verification; (3) Weakness Analysis and Instructional Guidance, in which scored-response patterns are clustered across cohorts to surface systematic knowledge gaps and generate targeted remediation strategies alongside instructor-facing briefings; (4) Peer-Review Triage Assistant, which processes incoming review-invitation manuscripts to produce concise preview briefs—covering structural completeness, novelty cross-referencing, methodological sanity checks, and reviewer-load recommendations—enabling rapid accept/decline/redirect decisions; and (5) Bilingual Content Delivery, generating Traditional Chinese + English slides, speaker notes, and multilingual narration via zero-shot voice cloning (Qwen3-TTS) from a 15-second reference sample. We report empirical results including the evaluation of ~70 theses per semester, processing of ~500 weekly quiz responses with automated scoring and psychometric analysis, cohort-level weakness identification across three courses, preview briefs for ~30 review-invitation manuscripts per month, and 86 slide-level narrations for a Bayesian Optimization course. We discuss critical deployment lessons—ARM64 compatibility, unified memory trade-offs, thermal management, PDF extraction edge cases, and the ethical considerations of AI-assisted peer review—alongside benefits of IP preservation, Traditional Chinese localization, and deployment democratization at one-third the cost of conventional workstations. The talk concludes with a proposed "signal triage" architecture where local AI handles 80% of routine patterns, enabling scholars to focus on the 20% that demand deep human judgment, offering a reproducible template for budget-constrained institutions pursuing privacy-preserving academic AI at lab scale.

 

Assoc. Prof. Fabrizio Marozzo

University of Calabria, Italy

Fabrizio Marozzo is an Associate Professor in Computer Engineering at the DIMES Department, University of Calabria, Italy, since 2019. He is also co-founder of DtoK Lab S.r.l., an academic spin-off of the University of Calabria established in 2014.
His current research explores big data analytics, generative artificial intelligence, large language models, machine learning, and distributed computing architectures spanning the cloud and edge continuum.
He is an IEEE Senior Member and serves as Associate Editor for several international journals, including IEEE Access, IEEE Transactions on Big Data, Journal of Big Data (Springer), SN Computer Science (Springer), Scalable Computing: Practice and Experience, Big Data and Cognitive Computing (MDPI), Algorithms (MDPI), and Heliyon (Elsevier). He has also served on the Program Committees of leading international conferences such as the IEEE International Conference on Big Data (IEEE BigData), International Joint Conference on Artificial Intelligence (IJCAI), IEEE International Conference on Data Mining (ICDM), Pacific-Asia Conference on Knowledge Discovery and Data Mining (PAKDD), and the European Conference on Parallel and Distributed Computing (Euro-Par).
He has published about 150 scientific papers, including the books Data Analysis in the Cloud and Programming Big Data Applications: Scalable Tools and Frameworks for Your Needs.

Title:  When AI Becomes a Bridge Across Expertise: Human-AI Collaboration and the Limits of Agentic AI
Abstract: This talk discusses AI not only as a productivity tool or an autonomous agent, but as a bridge across expertise. AI can help non-specialists engage with complex technical domains, connect knowledge from different experts, and support multi-step problem solving. At the same time, these capabilities still depend on human judgment, expert collaboration, and institutional processes. The talk also highlights some important limits of current agentic AI systems in real-world environments, especially in situations that require trust, adaptation, and coordination. The main argument is that the future of AI lies in effective and accountable human–AI collaboration rather than in the replacement of human agency.

 

Assoc. Prof. Moncef Garouani

Université Toulouse Capitole, France

Moncef GAROUANI is an Associate Professor at the University Toulouse Capitole and a member of the IRIT lab (SIG team) since 2023. He also serves as the Head of International Relations for the Faculty of Computer Science at University Toulouse Capitole. In 2022, Dr. Garouani successfully completed his doctoral studies, earning a Ph.D. in Computer Science through a collaborative program between Littoral Côte d'Opale University in France and Hassan II University of Casablanca in Morocco. His research interests lie at the intersection of several key areas within AI, including Automated Machine Learning (AutoML), Explainable AI (XAI), and meta-learning. He focuses on developing sophisticated techniques for the automatic selection and optimization of machine learning algorithms, as well as investigating methods to make the decision-making processes of AI models more understandable. Beyond these core areas, Dr. Garouani's research also encompasses Natural Language Processing (NLP) and Computer Vision, exploring how AI can be leveraged to understand and interpret diverse data modalities.

 

Prof. Anand Nayyar

Duy Tan University, Vietnam

Dr. Anand Nayyar received Ph.D (Computer Science) from Desh Bhagat University in 2017 in the area of Wireless Sensor Networks, Swarm Intelligence and Network Simulation. He is currently working in School of Computer Science and Artificial Intelligence (SCA)-Duy Tan University, Da Nang, Vietnam as Professor, Scientist, Vice-Chairman (Research) and Director- IoT and Intelligent Systems Lab. A Certified Professional with 280+ Professional certifications from CISCO, Microsoft, CompTIA, Amazon, Alibaba Cloud, Oracle, Google, Salesforce, Tableau, FinOps, Beingcert, EXIN, GAQM, Cyberoam and many more. Published more than 300+ Research Papers in various High-Quality ISI-SCI/SCIE/SSCI Impact Factor- Q1, Q2, Q3, Q4 Journals cum Scopus/ESCI indexed Journals, 80+ Papers in International Conferences indexed with Springer, IEEE and ACM Digital Library, 60+ Book Chapters in various SCOPUS/WEB OF SCIENCE Indexed Books with Springer, CRC Press, Wiley, IET, Elsevier with Citations: (Google Scholar): 23600+, H-Index: 78 and I-Index: 322; (Scopus): 13600+; H-index: 61. Member of more than 60+ Associations as Senior and Life Member like: IEEE (Senior Member) and ACM (Senior Member). He has authored/co-authored cum Edited 70+ Books of Computer Science. Associated with more than 600+ International Conferences as Programme Committee/Chair/Advisory Board/Review Board member. He has completed 1 Grassroot and 1 ASEAN Project. He has 18 Australian Patents, 16 German Patents, 4 Japanese Patents, 44 Indian Design cum Utility Patents, 13 UK Patents, 1 USA Patent, 3 Indian Copyrights and 2 Canadian Copyrights to his credit in the area of Wireless Communications, Artificial Intelligence, Cloud Computing, IoT, Healthcare, Drones, Robotics and Image Processing. He has guided more than 200+ Undergraduate Students (B.S. Degree), 30 MCA, 7 M.S. Students and completed 1 Ph.D Student and currently 3 Ph.D Scholars are working under him. He has completed 4 Research Grants Projects including 1 ASEAN and 1 Glocal 30 Project and 1 Grassroot Project in DTU. Awarded 58 Awards for Teaching and Research—Young Scientist, Best Scientist, Best Senior Scientist, Asia Top 50 Academicians and Researchers, Young Researcher Award, Outstanding Researcher Award, Excellence in Teaching, Best Senior Scientist Award, DTU Best Professor and Researcher Award- 2019, 2020-2021, 2022, 2022-2023, 2023-2024, 2024-2025, Distinguished Scientist Award by National University of Singapore, Obada Prize 2023, Lifetime Achievement Award 2023, 2024; Asian Admirable Achievers 2024; Distinguished Academic Leader 2024, Lifetime Achievement Award 2024 and many more.
He is listed in Top 2% Scientists as per Stanford University (2020, 2021, 2022, 2023, 2024, 2025), Ad Index (Rank No:1 Duy Tan University, Rank No:2 Computer Science in Viet Nam) and Listed on Research.com (Top Scientist of Computer Science in Viet Nam- National Ranking: 2; D-Index: 56; World Ranking: 3694).
He is acting as Associate Editor for Computer Communications (Elsevier), International Journal of Sensor Networks (IJSNET) (Inderscience), Tech Science Press- IASC, Cogent Engineering, Human Centric Computing and Information Sciences (HCIS), IEEE Transactions on Artificial Intelligence (IEEE TAI), Indonesian Journal of Electrical Engineering and Computer Science, IJFC, IJISP, IJDST, IJCINI, IJGC, IJSIR, IJBDCN, IJNR, IJSI, IJIES. He is acting as Managing Editor of IGI-Global Journal, USA titled “International Journal of Knowledge and Systems Science (IJKSS)”. He has reviewed more than 5700+ Articles for diverse Web of Science and Scopus Indexed Journals. He is currently researching in the area of Wireless Sensor Networks, Internet of Things, Swarm Intelligence, Cloud Computing, Artificial Intelligence, Drones, Blockchain, Cyber Security, Healthcare Informatics, Big Data and Wireless Communications.

Title:  From Pipeline to Impact: DataOps Best Practices and 9 Steps to Transform Data Science Delivery
Abstract: Modern data teams are under pressure to do two things at once: build reliable, automated data pipelines and turn data science work into measurable business impact. This keynote brings those two priorities together through a practical DataOps lens. Drawing on the themes of “Best Practices in DataOps: How to Create Agile, Automated Data Pipelines” and “DataOps: 9 Steps to Transform Your Data Science Impact,” the session explores how organizations can move beyond isolated analytics efforts toward a scalable, collaborative, and outcome-driven operating model. The keynote will show how DataOps helps bridge the long-standing gaps between data engineering, analytics, and data science by applying agile principles, automation, testing, monitoring, and continuous improvement to the full data lifecycle. It will examine the foundations of modern pipeline design, including reproducibility, observability, data quality controls, orchestration, versioning, and rapid iteration. At the same time, it will highlight the organizational shifts required to ensure that technical excellence actually translates into business value. A central focus of the session is the transformation of data science from a project-based function into a production-ready capability. Through a clear nine-step perspective, attendees will learn how to reduce friction between experimentation and deployment, improve cross-functional trust, accelerate feedback loops, and deliver insights that are timely, dependable, and actionable. Rather than treating DataOps as just another framework, this keynote presents it as a disciplined yet human-centered approach to working with data at scale. It is designed for leaders, engineers, analysts, and data scientists who want to create agile teams, automate with purpose, and build data products that are both technically robust and strategically meaningful. The result is a roadmap for turning better pipelines into better decisions—and better decisions into lasting impact.
 

Assoc. Prof. Paul S. Pang

Federation University, Australia

Dr. Pang is an Associate Professor of cyber security and he leads the Internet Commerce Security Lab (ICSL) at the Institute of Innovation, Science and Sustainability, Federation University Australia. Before joining Federation University, he was a Professor of Data Analytics and Director of Centre Computational Intelligence for Cybersecurity at the Unitec Institute of Technology, New Zealand. His research specializes in AI for cybersecurity, applied blockchain, and digital agricultural traceability. Dr. Pang is a Senior Member of IEEE, the Event Editor of Neural Network Journal Elsevier, Associate Editor for Springer's Pattern Analysis & Applications, and Advisory Board Member for the Journal of Cybersecurity Technology (Taylor & Francis).

Title: Structured Knowledge as the Backbone of Domain-Specific LLM Agents: A Harness Engineering Perspective
Abstract: Deploying large language models (LLMs) for real-world domain tasks requires more than capable models and well-crafted prompts. This talk introduces LLM harness engineering, a five-dimension framework that systematically governs how agents are built: context engineering, prompt engineering, knowledge base (KB) design, task decomposition, and task ordering. We argue that the knowledge base is the pivotal dimension, acting as the structured substrate from which reliable, grounded, and auditable agent behaviour emerges. We present two production-oriented case studies from the Australian agricultural sector. The first is an Apple Traceability Agent that queries a lifecycle knowledge base structured around a four-dimensional schema (Timing, Observation, Key Measurement, Management) with BBCH phenological codes as a cross-variety alignment mechanism. The second is an Agricultural Compliance Audit (ACA) Agent deployed on the GrainSupp blockchain platform, which decomposes the Australian Export Control (Plants and Plant Products) Rules 2021 into hierarchical audit checkpoints and applies section-specific tuned prompts against farm evidence. Evaluated across four LLMs, prompt-tuned agents achieve up to 99.1% compliance decision accuracy, a 12.8 percentage point improvement over untuned baselines at a cost of $0.031 per audit.

 

Assoc. Prof. Ramesh Kumar Ayyasamy

Universiti Tunku Abdul Rahman, Malaysia

Ramesh Kumar Ayyasamy (Senior Member, IEEE) is an Associate Professor in the Faculty of Information and Communication Technology at Universiti Tunku Abdul Rahman (UTAR), Malaysia. Dr Ramesh holds a Ph.D. in Information Technology from Monash University, Australia, which he completed in 2013. He has over 23 years of teaching and research experience in Computer Science and Information Systems. He has held various academic and research roles at multiple institutions throughout his career. Dr. Ramesh's research expertise lies in AI-driven text analytics, focusing on sentiment analysis, deep learning for healthcare imaging, and semantic image segmentation. His work bridges theoretical foundations and practical applications, contributing to natural language processing, computer vision, smart city development, and health informatics. In addition to his research activities, he plays an active role in the academic community as a reviewer for leading WoS/Scopus ranked international journals and IEEE indexed conferences. He serves on the editorial boards of several scholarly journals, including Discover Artificial Intelligence, Journal of Engineering and Applied Science, International Journal of Innovative Engineering, Science and Technology and Iris Online Journal of Cyber security & Computing technology.

Title: Architecting the Intelligent Future: Data, Computing, and Society
Abstract: The rapid growth of big data and advanced computing technologies is reshaping the foundations of modern society. From healthcare and finance to education, manufacturing, and smart cities, intelligent data-driven systems now influence how organizations operate, make decisions, and create value. This keynote explores the evolving relationship between big data, artificial intelligence, cloud computing, and high-performance computing in building the next generation of digital ecosystems. Beyond technological advancement, the keynote addresses critical challenges in data governance, cybersecurity, privacy, ethics, and sustainability. As intelligent systems become deeply integrated into everyday life, responsible innovation and human-centered computing must remain central priorities. By connecting technological progress with societal impact, this keynote provides a forward-looking perspective on how researchers, industry leaders, and policymakers can collaboratively architect an intelligent, secure, and inclusive digital future driven by big data and advanced computing.

 

Assoc. Prof. Quazi Mamun

Charles Sturt University, Australia

Dr Quazi Mamun is an Associate Professor in Computing and the Higher Degree by Research (HDR) Coordinator at Charles Sturt University, possessing over 25 years of experience in teaching, research, and academic leadership. His research expertise focuses on artificial intelligence, cybersecurity, and the Internet of Things, with an emphasis on creating trustworthy digital systems through international collaboration. A Senior Member of the IEEE, Dr Mamun has published more than 90 research papers in top-tier international journals and conferences and is recognised among the top 1% of European Alliance for Innovation (EAI) members for his significant contributions to the global research community. He has successfully led numerous interdisciplinary projects, including a prestigious $5,000,000 ASPIRE grant from the Japan Science and Technology Agency to investigate infrastructure for ambient intelligence. Dr Mamun earned his PhD from Monash University, following a Master of Science from Waseda University in Japan and a Bachelor of Science from the Bangladesh University of Engineering and Technology. His academic influence extends globally through visiting professorships in India and collaborative initiatives across Asia, Europe, and North America. In addition to his research impact, he is an expert in curriculum renewal and has supervised 10 HDR students, demonstrating a long-standing commitment to high-quality postgraduate research training and the mentorship of early-career researchers.

Title: Towards Trustworthy Intelligent Systems: A Research Journey Across Big Data, AI, Cybersecurity and IoT
Abstract: The rapid growth of big data, artificial intelligence, cybersecurity technologies and the Internet of Things is reshaping how intelligent systems are designed, deployed and governed. While these technologies offer significant opportunities for automation, prediction, optimisation and decision support, they also introduce new challenges related to trust, security, privacy, transparency, resilience and responsible use. This invited talk presents a research journey across several interconnected areas of computing, including big data analytics, AI-enabled cybersecurity, secure IoT systems, blockchain-assisted trust management, cloud-native compliance automation and intelligent cyber-physical systems.
Rather than focusing on a single project, the presentation reflects on a broader research programme aimed at developing trustworthy intelligent systems for real-world digital environments. It discusses how data-driven techniques can be used to detect threats, support secure decision-making, improve system resilience and enable scalable intelligence across distributed infrastructures. The talk also highlights the role of emerging approaches such as federated learning, edge intelligence, zero-trust architectures, blockchain-enabled verification and agentic AI in addressing contemporary challenges in connected and autonomous environments.

Drawing on research experiences across IoT security, communication networks, autonomous systems, smart infrastructure, healthcare data management, cloud compliance and cybersecurity awareness, the presentation identifies key lessons for designing intelligent systems that are not only accurate and efficient but also secure, explainable, accountable and socially responsible. The talk concludes by outlining future research directions for trustworthy intelligent computing, with emphasis on interdisciplinary collaboration, human-centred design, responsible AI governance and the translation of research outcomes into practical impact.

 

Assoc. Prof. Marshima Mohd Rosli

Universiti Teknologi MARA, Malaysia

Marshima Mohd Rosli is an Associate Professor in the Department of Computing Sciences at the Faculty of Computer and Mathematical Sciences, Universiti Teknologi MARA, Malaysia. She holds a PhD in Computer Science from the University of Auckland, New Zealand. Her research primarily focuses on software engineering, system development, artificial intelligence, and data analytics. Her research primarily focuses on improving software and data quality through machine learning, empirical evaluation, and intelligent systems. A passionate advocate for multidisciplinary innovation, she serves as a researcher at the Cardiovascular Advancement and Research Excellence (CARE) Institute, Universiti Teknologi MARA, where she collaborates with clinical experts to develop AI-driven healthcare solutions. She has contributed numerous publications to renowned conferences and journals such as IEEE and ACM.

Title: Beyond the Black Box: A Data-Driven Approach to Optimizing Active Learning Strategies in Healthcare
Abstract: Data-driven models are heavily utilized in domains like healthcare for precise predictions, but they require massive volumes of high-quality labeled data. Because acquiring labeled medical data is notoriously expensive and constrained by privacy issues, Active Learning (AL) has emerged as a crucial methodology to minimize labeling costs by selectively querying only the most informative samples. However, the efficacy of AL is highly vulnerable to data quality issues, such as missing values, class imbalance, high dimensionality, outliers, and data redundancy. Currently, many researchers treat AL strategies as a black box, applying them without systematically analyzing the underlying dataset characteristics, which often results in unpredictable or suboptimal predictive performance.This presentation introduces a novel data-driven selection model designed to solve the no free lunch problem in active learning. By conducting extensive experimental evaluations across various medical datasets, this study systematically mapped the relationships between specific data quality challenges and the performance of baseline AL strategies, including Uncertainty Sampling, Query-by-Committee (QBC), and Diversity Sampling.